HK1241135B - Audio coding method and related device - Google Patents

Description

Audio coding method and related device

Technical Field

The present invention relates to audio coding technology, and in particular to an audio coding method and related devices.

Background Art

For a long time in the past, the encoding of speech signals and the encoding of non-speech signals (such as music) were relatively independent, that is, the encoding of speech signals was implemented by a dedicated speech encoder, while the encoding of non-speech signals was implemented by a dedicated non-speech encoder (wherein the non-speech encoder can also be called a general audio encoder).

Speech encoders are generally not used to encode non-speech signals, and vice versa. This is not only because speech and non-speech signal encoding are relatively independent in coding theory, but also because these two types of signals are often relatively independent in practical applications. For example, in voice communication networks, since voice has long been the sole or primary signal source and bandwidth constraints are strict, various low-rate speech encoders are widely used in voice communication networks. In audio and video, entertainment, and other applications, non-speech encoders are widely used because non-speech signals make up the majority of the signal source and these applications have relatively high requirements for audio quality and relatively loose bitrate requirements.

In recent years, more and more multimedia sources, such as ringback tone (CRBT), have appeared in traditional voice communication networks. This places higher demands on the encoding quality of encoders. Dedicated voice encoders can no longer provide the high encoding quality required for these multimedia signals. New encoding technologies, such as hybrid audio encoders, have emerged.

A hybrid audio codec is an audio codec that includes both subcoders for encoding speech signals and subcoders for encoding non-speech signals. The hybrid audio codec dynamically selects the most suitable subcoder from among all subcoders to encode the input audio signal. Choosing the most suitable subcoder to encode the current input audio frame is a key function and requirement of the hybrid codec. This subcoder selection, also known as mode selection, directly impacts the encoding quality of the hybrid codec.

Existing technologies typically use a closed-loop mode to select sub-encoders. This means each sub-encoder is used to encode the current input audio frame once, and the optimal sub-encoder is selected by directly comparing the quality of the encoded current audio frame. However, the disadvantage of closed-loop selection is that it results in relatively high encoding complexity (because each sub-encoder is used to encode the current input audio frame once), which in turn increases the actual audio encoding overhead.

Summary of the Invention

The embodiments of the present invention provide an audio encoding method and related devices, in order to reduce the overhead of audio encoding.

A first aspect of an embodiment of the present invention provides an audio encoding method, including:

estimating a reference linear prediction efficiency for the current audio frame;

Determining an audio coding mode that matches a reference linear prediction efficiency of the current audio frame;

The current audio frame is audio-encoded according to an audio coding method that matches a reference linear prediction efficiency of the current audio frame.

In combination with the first aspect, in a first possible implementation of the first aspect,

The reference linear prediction efficiency includes at least one of the following linear prediction efficiencies: a reference long-term linear prediction efficiency, a reference short-term linear prediction efficiency, and a reference comprehensive linear prediction efficiency.

In combination with the first possible implementation of the first aspect, in a second possible implementation of the first aspect, the reference comprehensive linear prediction efficiency is the sum, weighted sum or average of the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency.

In conjunction with the first possible implementation manner of the first aspect, in a third possible implementation manner of the first aspect, if the reference linear prediction efficiency of the current audio frame includes a reference long-time linear prediction efficiency of the current audio frame and a reference short-time linear prediction efficiency of the current audio frame, then determining an audio coding mode that matches the reference linear prediction efficiency of the current audio frame includes:

If the reference long-term linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-term linear prediction efficiency of the current audio frame is less than a second threshold, determining that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode not based on linear prediction;

and/or,

If the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a second threshold, it is determined that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

In conjunction with the first possible implementation manner of the first aspect, in a fourth possible implementation manner of the first aspect, if the reference linear prediction efficiency of the current audio frame includes a reference long-term linear prediction efficiency of the current audio frame, then determining the audio coding mode that matches the reference linear prediction efficiency of the current audio frame includes:

If the reference long-term linear prediction efficiency of the current audio frame is greater than or equal to a third threshold, determining that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode based on linear prediction;

And/or, if the reference long-term linear prediction efficiency of the current audio frame is less than a fourth threshold, it is determined that the audio coding method matching the reference linear prediction efficiency of the current audio frame is an audio coding method not based on linear prediction.

In combination with the first possible implementation of the first aspect, in the fifth possible implementation of the first aspect, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, then determining the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: determining a first linear prediction efficiency interval in which the reference long-time linear prediction efficiency of the current audio frame falls, and determining a first audio encoding method having a mapping relationship with the first linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio encoding method based on linear prediction, wherein the first audio encoding method is an audio encoding method that matches the reference linear prediction efficiency of the current audio frame, and the first audio encoding method is an audio encoding method based on linear prediction or an audio encoding method not based on linear prediction.

In conjunction with the first possible implementation manner of the first aspect, in a sixth possible implementation manner of the first aspect, if the reference linear prediction efficiency of the current audio frame includes a reference short-time linear prediction efficiency of the current audio frame, then determining the audio coding mode that matches the reference linear prediction efficiency of the current audio frame includes:

If the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a fifth threshold, determining that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode based on linear prediction;

And/or, if the reference short-time linear prediction efficiency of the current audio frame is less than a fifth threshold, it is determined that the audio coding mode matching the reference linear prediction efficiency of the current audio frame is an audio coding mode not based on linear prediction.

In combination with the first possible implementation of the first aspect, in the seventh possible implementation of the first aspect, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then determining the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: determining a second linear prediction efficiency interval in which the reference short-time linear prediction efficiency of the current audio frame falls, and determining a second audio encoding method having a mapping relationship with the second linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio encoding method based on linear prediction, wherein the second audio encoding method is an audio encoding method that matches the reference linear prediction efficiency of the current audio frame, and the second audio encoding method is an audio encoding method based on linear prediction or an audio encoding method not based on linear prediction.

In combination with the first possible implementation manner of the first aspect or the second possible implementation manner of the first aspect, in an eighth possible implementation manner of the first aspect,

If the reference linear prediction efficiency of the current audio frame includes a reference integrated linear prediction efficiency of the current audio frame, determining an audio coding mode that matches the reference linear prediction efficiency of the current audio frame includes:

If the reference integrated linear prediction efficiency of the current audio frame is greater than or equal to a sixth threshold, determining that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode based on linear prediction;

And/or, if the reference comprehensive linear prediction efficiency of the current audio frame is less than a sixth threshold, it is determined that the audio encoding method matching the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

In combination with the first possible implementation manner of the first aspect or the second possible implementation manner of the first aspect, in a ninth possible implementation manner of the first aspect,

If the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then determining the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: determining a third linear prediction efficiency interval into which the reference comprehensive linear prediction efficiency of the current audio frame falls, and determining a third audio encoding method that has a mapping relationship with the third linear prediction efficiency interval based on a mapping relationship between linear prediction efficiency intervals and audio encoding methods based on linear prediction, wherein the third audio encoding method is an audio encoding method that matches the reference linear prediction efficiency of the current audio frame, and the third audio encoding method is an audio encoding method based on linear prediction or an audio encoding method not based on linear prediction.

In combination with the first to ninth possible implementations of the first aspect, in a tenth possible implementation of the first aspect, the reference long-term linear prediction efficiency of the current audio frame is estimated by: estimating the long-term linear prediction efficiency of the current audio frame, wherein the long-term linear prediction efficiency of the current audio frame is the reference long-term linear prediction efficiency of the current audio frame; or

The reference long-time linear prediction efficiency of the current audio frame is estimated by: estimating the long-time linear prediction efficiency of the current audio frame; obtaining the linear prediction efficiencies of N1 historical audio frames of the current audio frame; calculating a first statistical value of the linear prediction efficiencies of the N1 historical audio frames and the long-time linear prediction efficiency of the current audio frame, wherein N1 is a positive integer, and the first statistical value is the reference long-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N11 historical audio frames is at least one of the following linear prediction efficiencies: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency; the N11 historical audio frames are a subset of the N1 historical audio frames; or,

The reference long-time linear prediction efficiency of the current audio frame is estimated by: estimating the long-time linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiencies of N2 historical audio frames of the current audio frame; calculating the reference linear prediction efficiencies of the N2 historical audio frames and a second statistical value of the long-time linear prediction efficiency of the current audio frame, wherein N2 is a positive integer, and the second statistical value is the reference long-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each of the N21 historical audio frames is at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, and the N21 historical audio frames are a subset of the N2 historical audio frames; or,

The reference long-term linear prediction efficiency of the current audio frame is estimated by: estimating the long-term linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiencies of N4 historical audio frames of the current audio frame, and obtaining the linear prediction efficiencies of N3 historical audio frames of the current audio frame; and calculating a third statistical value of the linear prediction efficiencies of the N3 historical audio frames, the reference linear prediction efficiencies of the N4 historical audio frames, and the long-term linear prediction efficiency of the current audio frame, wherein N3 and N4 are positive integers, and the third statistical value is the reference long-term linear prediction efficiency of the current audio frame. , the linear prediction efficiency of each of the N31 historical audio frames is at least one of the following linear prediction efficiency: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; wherein, the reference linear prediction efficiency of each of the N41 historical audio frames is at least one of the following linear prediction efficiency: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the N31 historical audio frames are a subset of the N3 historical audio frames, and wherein the N41 historical audio frames are a subset of the N4 historical audio frames.

In combination with the first to ninth possible implementations of the first aspect, in an eleventh possible implementation of the first aspect,

The reference short-time linear prediction efficiency of the current audio frame is estimated by: estimating the short-time linear prediction efficiency of the current audio frame, wherein the short-time linear prediction efficiency of the current audio frame is the reference short-time linear prediction efficiency of the current audio frame; or

The reference short-time linear prediction efficiency of the current audio frame is estimated by: estimating the short-time linear prediction efficiency of the current audio frame; obtaining the linear prediction efficiencies of N5 historical audio frames of the current audio frame; calculating a fourth statistical value of the linear prediction efficiencies of the N5 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein N5 is a positive integer, and the fourth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N51 historical audio frames is at least one of the following linear prediction efficiencies: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency, and the N51 historical audio frames are a subset of the N5 historical audio frames; or,

The reference short-time linear prediction efficiency of the current audio frame is estimated in the following manner: estimating the short-time linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiencies of N6 historical audio frames of the current audio frame; calculating the fifth statistical value of the reference linear prediction efficiencies of the N6 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein N6 is a positive integer, and the fifth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each historical audio frame in the N61 historical audio frames is at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the N61 historical audio frames are a subset of the N6 historical audio frames; or,

The reference short-time linear prediction efficiency of the current audio frame is estimated by: estimating the short-time linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiency of N8 historical audio frames of the current audio frame; obtaining the linear prediction efficiency of N7 historical audio frames of the current audio frame; calculating the sixth statistical value of the linear prediction efficiency of the N7 historical audio frames, the reference linear prediction efficiency of the N8 historical audio frames, and the short-time linear prediction efficiency of the current audio frame, wherein N7 and N8 are positive integers, and the sixth statistical value is the reference short-time linear prediction efficiency of the current audio frame. The linear prediction efficiency of each historical audio frame in the N71 historical audio frames is at least one of the following linear prediction efficiency: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; the reference linear prediction efficiency of each historical audio frame in the N81 historical audio frames is at least one of the following linear prediction efficiency: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the N71 historical audio frames are a subset of the N7 historical audio frames, and the N81 historical audio frames are a subset of the N8 historical audio frames.

In combination with the eleventh possible implementation of the first aspect, in the twelfth possible implementation of the first aspect, the estimating the short-time linear prediction efficiency of the current audio frame includes: obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame.

In conjunction with the twelfth possible implementation manner of the first aspect, in a thirteenth possible implementation manner of the first aspect, obtaining the short-term linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame includes:

Calculate an energy change rate of a current audio frame before and after short-time linear prediction is performed, where the energy change rate is a short-time linear prediction efficiency of the current audio frame, or the short-time linear prediction efficiency of the current audio frame is obtained by transforming the energy change rate, and the energy of the current audio frame after short-time linear prediction is an energy of a linear prediction residual of the current audio frame.

In combination with the thirteenth possible implementation of the first aspect, in the fourteenth possible implementation of the first aspect, the energy change rate of the current audio frame before and after short-time linear prediction is the ratio of the energy of the current audio frame before short-time linear prediction to the energy of the linear prediction residual of the current audio frame.

In combination with the tenth possible implementation manner of the first aspect, in a fifteenth possible implementation manner of the first aspect,

The estimating of the long-term linear prediction efficiency of the current audio frame includes: obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal, wherein the correlation is the long-term linear prediction efficiency of the current audio frame, or the long-term linear prediction efficiency of the current audio frame is obtained based on the correlation, wherein the first historical linear prediction signal is the first historical linear prediction excitation or the first historical linear prediction residual; the first historical linear prediction residual is the linear prediction residual of the historical audio frame of the current audio frame, and the first historical linear prediction excitation is the linear prediction excitation of the historical audio frame of the current audio frame.

In combination with the fifteenth possible implementation manner of the first aspect, in a sixteenth possible implementation manner of the first aspect, obtaining, based on the linear prediction residual of the current audio frame and the first historical linear prediction signal, a correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal includes:

Calculating the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal;

or,

multiplying a linear prediction residual of a current audio frame by a gain factor to obtain a gain linear prediction residual of the current audio frame, and calculating a correlation between the gain linear prediction residual of the current audio frame and a first historical linear prediction signal, wherein the calculated correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal;

Alternatively, the first historical linear prediction signal is multiplied by a gain factor to obtain a first historical linear prediction signal after gain, and the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal after gain is calculated, wherein the calculated correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal after gain is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

In combination with the fifteenth possible implementation of the first aspect or the sixteenth possible implementation of the first aspect, in the seventeenth possible implementation of the first aspect, the first historical linear prediction excitation or the first historical linear prediction residual is determined based on the fundamental tone of the current audio frame.

In combination with the fifteenth to seventeenth possible implementations of the first aspect, in an eighteenth possible implementation of the first aspect, a time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to a time-domain correlation between other historical linear prediction excitations and the linear prediction residual of the current audio frame;

Alternatively, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction residuals and the linear prediction residual of the current audio frame.

In combination with the fifteenth to eighteenth possible implementations of the first aspect, in the nineteenth possible implementation of the first aspect, the first historical linear prediction excitation is a linear prediction excitation generated by audio encoding the historical audio frames of the current audio frame using a linear prediction-based encoding method.

In combination with the fifteenth to nineteenth possible implementations of the first aspect, in the twentieth possible implementation of the first aspect, the first historical linear prediction residual is obtained based on the time domain signal of the first historical audio frame of the current audio frame and the linear prediction coefficient of the first historical audio frame, wherein the linear prediction coding coefficient of the first historical audio frame is a quantized linear prediction coefficient or an unquantized linear prediction coefficient.

In combination with the fifteenth to twentieth possible implementations of the first aspect, in the twenty-first possible implementation of the first aspect, the linear prediction residual of the current audio frame is obtained based on the time domain signal of the current audio frame and the linear prediction coefficient of the current audio frame, wherein the linear prediction coefficient of the current audio frame is a quantized linear prediction coefficient or an unquantized linear prediction coefficient.

In combination with the fifteenth to twenty-first possible implementations of the first aspect, in the twenty-second possible implementation of the first aspect, the first historical linear prediction excitation is a superimposed excitation of an adaptive codebook excitation and a fixed codebook excitation, or the first historical linear prediction excitation is an adaptive codebook excitation.

In combination with the fifteenth to twenty-second possible implementations of the first aspect, in the twenty-third possible implementation of the first aspect, the correlation is the cross-correlation function value in the time domain and/or the cross-correlation function value in the frequency domain, or the correlation is the distortion in the time domain and/or the distortion in the frequency domain.

In combination with the twenty-three possible implementations of the first aspect, in the twenty-fourth possible implementation of the first aspect, the distortion in the frequency domain is the sum or weighted sum of the distortions of K1 frequency points in the frequency domain, or the distortion in the frequency domain is the sum or weighted sum of the distortions on K2 subbands in the frequency domain, and K1 and K2 are positive integers.

In combination with the twenty-four possible implementations of the first aspect, in a twenty-fifth possible implementation of the first aspect, the weighting coefficient corresponding to the weighted sum of the distortion values is a perceptual weighting coefficient reflecting a psychoacoustic model.

A second aspect of an embodiment of the present invention provides an audio encoder, including:

an estimating unit, configured to estimate a reference linear prediction efficiency of a current audio frame;

a determining unit, configured to determine an audio coding mode that matches the reference linear prediction efficiency of the current audio frame estimated by the estimating unit;

An encoding unit is configured to perform audio encoding on the current audio frame according to the audio encoding mode determined by the determining unit and matching the reference linear prediction efficiency of the current audio frame.

In combination with the second aspect, in a first possible implementation of the second aspect, the reference linear prediction efficiency includes at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency.

In combination with the first possible implementation of the second aspect, in a second possible implementation of the second aspect, the reference comprehensive linear prediction efficiency is the sum, weighted sum or average of the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency.

In conjunction with the first possible implementation manner of the second aspect, in a third possible implementation manner of the second aspect, if the reference linear prediction efficiency of the current audio frame includes a reference long-time linear prediction efficiency of the current audio frame and a reference short-time linear prediction efficiency of the current audio frame, the determining unit is specifically configured to:

If the reference long-term linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-term linear prediction efficiency of the current audio frame is less than a second threshold, determining that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode not based on linear prediction;

and/or,

If the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a second threshold, it is determined that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

In combination with the first possible implementation manner of the second aspect, in a fourth possible implementation manner of the second aspect, if the reference linear prediction efficiency of the current audio frame includes a reference long-term linear prediction efficiency of the current audio frame, the determining unit is specifically configured to:

If the reference long-term linear prediction efficiency of the current audio frame is greater than or equal to a third threshold, determining that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode based on linear prediction;

And/or, if the reference long-term linear prediction efficiency of the current audio frame is less than a fourth threshold, it is determined that the audio coding method matching the reference linear prediction efficiency of the current audio frame is an audio coding method not based on linear prediction.

In combination with the first possible implementation of the second aspect, in the fifth possible implementation of the second aspect, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, the determination unit is specifically used to: determine the first linear prediction efficiency interval into which the reference long-time linear prediction efficiency of the current audio frame falls, and determine a first audio coding method having a mapping relationship with the first linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio coding method based on linear prediction, wherein the first audio coding method is an audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the first audio coding method is an audio coding method based on linear prediction or an audio coding method not based on linear prediction.

In combination with the first possible implementation manner of the second aspect, in a sixth possible implementation manner of the second aspect, if the reference linear prediction efficiency of the current audio frame includes a reference short-time linear prediction efficiency of the current audio frame, the determining unit is specifically configured to:

If the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a fifth threshold, determining that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode based on linear prediction;

And/or, if the reference short-time linear prediction efficiency of the current audio frame is less than a fifth threshold, it is determined that the audio coding mode matching the reference linear prediction efficiency of the current audio frame is an audio coding mode not based on linear prediction.

In combination with the first possible implementation of the second aspect, in the seventh possible implementation of the second aspect, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, the determination unit is specifically used to: determine the second linear prediction efficiency interval into which the reference short-time linear prediction efficiency of the current audio frame falls, and determine a second audio coding method having a mapping relationship with the second linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio coding method based on linear prediction, wherein the second audio coding method is an audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the second audio coding method is an audio coding method based on linear prediction or an audio coding method not based on linear prediction.

In combination with the first possible implementation manner of the second aspect or the second possible implementation manner of the second aspect, in an eighth possible implementation manner of the second aspect,

If the reference linear prediction efficiency of the current audio frame includes a reference integrated linear prediction efficiency of the current audio frame, the determining unit is specifically configured to:

If the reference integrated linear prediction efficiency of the current audio frame is greater than or equal to a sixth threshold, determining that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode based on linear prediction;

And/or, if the reference comprehensive linear prediction efficiency of the current audio frame is less than a sixth threshold, it is determined that the audio encoding method matching the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

In combination with the first possible implementation manner of the second aspect or the second possible implementation manner of the second aspect, in a ninth possible implementation manner of the second aspect,

If the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, the determination unit is specifically used to: determine a third linear prediction efficiency interval into which the reference comprehensive linear prediction efficiency of the current audio frame falls, and determine a third audio coding method having a mapping relationship with the third linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio coding method based on linear prediction, wherein the third audio coding method is an audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the third audio coding method is an audio coding method based on linear prediction or an audio coding method not based on linear prediction.

In combination with the first to ninth possible implementations of the second aspect, in a tenth possible implementation of the second aspect, in terms of estimating the reference long-term linear prediction efficiency of the current audio frame, the estimation unit is specifically configured to: estimate the long-term linear prediction efficiency of the current audio frame, wherein the long-term linear prediction efficiency of the current audio frame is the reference long-term linear prediction efficiency of the current audio frame; or

In terms of estimating the reference long-time linear prediction efficiency of the current audio frame, the estimation unit is specifically used to: estimate the long-time linear prediction efficiency of the current audio frame; obtain the linear prediction efficiencies of N1 historical audio frames of the current audio frame; calculate the linear prediction efficiencies of the N1 historical audio frames and a first statistical value of the long-time linear prediction efficiency of the current audio frame, wherein N1 is a positive integer, and the first statistical value is the reference long-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N11 historical audio frames is at least one of the following linear prediction efficiencies: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency; the N11 historical audio frames are a subset of the N1 historical audio frames; or,

In terms of estimating the reference long-time linear prediction efficiency of the current audio frame, the estimation unit is specifically used to: estimate the long-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiencies of N2 historical audio frames of the current audio frame; calculate the reference linear prediction efficiencies of the N2 historical audio frames and a second statistical value of the long-time linear prediction efficiency of the current audio frame, wherein N2 is a positive integer, wherein the second statistical value is the reference long-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each historical audio frame in the N21 historical audio frames is at least one of the following linear prediction efficiencies: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the N21 historical audio frames are a subset of the N2 historical audio frames; or,

In terms of estimating the reference long-time linear prediction efficiency of the current audio frame, the estimation unit is specifically used to: estimate the long-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of N4 historical audio frames of the current audio frame, and obtain the linear prediction efficiency of N3 historical audio frames of the current audio frame; calculate the linear prediction efficiency of the N3 historical audio frames, the reference linear prediction efficiency of the N4 historical audio frames, and the third statistical value of the long-time linear prediction efficiency of the current audio frame, wherein N3 and N4 are positive integers, and the third statistical value is the reference long-time linear prediction efficiency of the current audio frame. The linear prediction efficiency of each historical audio frame in the N31 historical audio frames is at least one of the following linear prediction efficiency: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; wherein, the reference linear prediction efficiency of each historical audio frame in the N41 historical audio frames is at least one of the following linear prediction efficiency: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the N31 historical audio frames are a subset of the N3 historical audio frames, and the N41 historical audio frames are a subset of the N4 historical audio frames.

In combination with the first to ninth possible implementations of the second aspect, in an eleventh possible implementation of the second aspect,

In terms of estimating the reference short-time linear prediction efficiency of the current audio frame, the estimating unit is specifically configured to: estimate the short-time linear prediction efficiency of the current audio frame, wherein the short-time linear prediction efficiency of the current audio frame is the reference short-time linear prediction efficiency of the current audio frame;

or,

In terms of estimating the reference short-time linear prediction efficiency of the current audio frame, the estimation unit is specifically used to: estimate the short-time linear prediction efficiency of the current audio frame; obtain the linear prediction efficiencies of N5 historical audio frames of the current audio frame; calculate a fourth statistical value of the linear prediction efficiencies of the N5 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein N5 is a positive integer, and the fourth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N51 historical audio frames is at least one of the following linear prediction efficiencies: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency, and the N51 historical audio frames are a subset of the N5 historical audio frames; or,

In terms of estimating the reference short-time linear prediction efficiency of the current audio frame, the estimation unit is specifically used to: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiencies of N6 historical audio frames of the current audio frame; calculate the fifth statistical value of the reference linear prediction efficiencies of the N6 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein N6 is a positive integer, and the fifth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each historical audio frame in the N61 historical audio frames is at least one of the following linear prediction efficiencies: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the N61 historical audio frames are a subset of the N6 historical audio frames; or,

In terms of estimating the reference short-time linear prediction efficiency of the current audio frame, the estimation unit is specifically used to: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of N8 historical audio frames of the current audio frame; obtain the linear prediction efficiency of N7 historical audio frames of the current audio frame; calculate the sixth statistical value of the linear prediction efficiency of the N7 historical audio frames, the reference linear prediction efficiency of the N8 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein N7 and N8 are positive integers, and the sixth statistical value is the reference short-time linear prediction efficiency of the current audio frame. The linear prediction efficiency of each historical audio frame in the N71 historical audio frames is at least one of the following linear prediction efficiency: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; the reference linear prediction efficiency of each historical audio frame in the N81 historical audio frames is at least one of the following linear prediction efficiency: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the N71 historical audio frames are a subset of the N7 historical audio frames, and the N81 historical audio frames are a subset of the N8 historical audio frames.

In combination with the eleventh possible implementation of the second aspect, in the twelfth possible implementation of the second aspect, in terms of estimating the short-time linear prediction efficiency of the current audio frame, the estimation unit is specifically used to: obtain the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame.

In combination with the twelfth possible implementation of the second aspect, in the thirteenth possible implementation of the second aspect, in the aspect of obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame, the estimation unit is specifically used to: calculate the energy change rate of the current audio frame before and after short-time linear prediction, wherein the energy change rate is the short-time linear prediction efficiency of the current audio frame, or the short-time linear prediction efficiency of the current audio frame is obtained based on the transformation of the energy change rate, wherein the energy of the current audio frame after short-time linear prediction is the energy of the linear prediction residual of the current audio frame.

In combination with the thirteenth possible implementation of the second aspect, in the fourteenth possible implementation of the second aspect, the energy change rate of the current audio frame before and after the short-time linear prediction is the ratio of the energy of the current audio frame before the short-time linear prediction to the energy of the linear prediction residual of the current audio frame.

In combination with the tenth possible implementation manner of the second aspect, in a fifteenth possible implementation manner of the second aspect,

In terms of estimating the long-term linear prediction efficiency of the current audio frame, the estimation unit is specifically used to: obtain the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the calculated linear prediction residual of the current audio frame and the first historical linear prediction signal, wherein the correlation is the long-term linear prediction efficiency of the current audio frame, or the long-term linear prediction efficiency of the current audio frame is obtained based on the correlation, wherein the first historical linear prediction signal is the first historical linear prediction excitation or the first historical linear prediction residual, the first historical linear prediction residual is the linear prediction residual of the historical audio frame of the current audio frame, and the first historical linear prediction excitation is the linear prediction excitation of the historical audio frame of the current audio frame.

In conjunction with the fifteenth possible implementation manner of the second aspect, in a sixteenth possible implementation manner of the second aspect, in the aspect of obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on calculating the linear prediction residual of the current audio frame and the first historical linear prediction signal, the estimation unit is specifically configured to: calculate the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal;

Alternatively, multiplying a linear prediction residual of a current audio frame by a gain factor to obtain a gain linear prediction residual of the current audio frame, and calculating a correlation between the gain linear prediction residual of the current audio frame and a first historical linear prediction signal, wherein the calculated correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal;

Alternatively, the first historical linear prediction signal is multiplied by a gain factor to obtain a first historical linear prediction signal after gain, and the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal after gain is calculated, wherein the calculated correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal after gain is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

In combination with the fifteenth possible implementation of the second aspect or the sixteenth possible implementation of the second aspect, in the seventeenth possible implementation of the second aspect, the first historical linear prediction excitation or the first historical linear prediction residual is determined based on the fundamental tone of the current audio frame.

In combination with the fifteenth to seventeenth possible implementations of the second aspect, in an eighteenth possible implementation of the second aspect, a time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to a time-domain correlation between other historical linear prediction excitations and the linear prediction residual of the current audio frame;

Alternatively, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction residuals and the linear prediction residual of the current audio frame.

In combination with the fifteenth to eighteenth possible implementations of the second aspect, in the nineteenth possible implementation of the second aspect, the first historical linear prediction excitation is a linear prediction excitation generated by audio encoding the historical audio frames of the current audio frame using a linear prediction-based encoding method.

In combination with the fifteenth to nineteenth possible implementations of the second aspect, in the twentieth possible implementation of the second aspect, the first historical linear prediction residual is obtained based on the time domain signal of the first historical audio frame of the current audio frame and the linear prediction coefficient of the first historical audio frame, wherein the linear prediction coding coefficient of the first historical audio frame is a quantized linear prediction coefficient or an unquantized linear prediction coefficient.

In combination with the fifteenth to twentieth possible implementations of the second aspect, in the twenty-first possible implementation of the second aspect, the linear prediction residual of the current audio frame is obtained based on the time domain signal of the current audio frame and the linear prediction coefficient of the current audio frame, wherein the linear prediction coefficient of the current audio frame is a quantized linear prediction coefficient or an unquantized linear prediction coefficient.

In combination with the fifteenth to twenty-first possible implementations of the second aspect, in the twenty-second possible implementation of the second aspect, the first historical linear prediction excitation is a superimposed excitation of an adaptive codebook excitation and a fixed codebook excitation, or the first historical linear prediction excitation is an adaptive codebook excitation.

In combination with the fifteenth to twenty-second possible implementations of the second aspect, in the twenty-third possible implementation of the second aspect, the correlation is the cross-correlation function value in the time domain and/or the cross-correlation function value in the frequency domain, or the correlation is the distortion in the time domain and/or the distortion in the frequency domain.

In combination with the twenty-three possible implementations of the second aspect, in the twenty-fourth possible implementation of the second aspect, the distortion in the frequency domain is the sum or weighted sum of the distortions of K1 frequency points in the frequency domain, or the distortion in the frequency domain is the sum or weighted sum of the distortions on K2 subbands in the frequency domain, and K1 and K2 are positive integers.

In combination with the twenty-four possible implementations of the second aspect, in a twenty-fifth possible implementation of the second aspect, the weighting coefficient corresponding to the weighted sum of the distortion values is a perceptual weighting coefficient reflecting a psychoacoustic model.

It can be seen that in the technical solutions of some embodiments of the present invention, the reference linear prediction efficiency of the current audio frame is first estimated; the audio coding method that matches the current audio frame is determined by the estimated reference linear prediction efficiency of the current audio frame, and the current audio frame is audio-encoded according to the determined audio coding method that matches the current audio frame. Since the above-mentioned solution does not need to perform the operation of completely encoding the current audio frame using each audio coding method required by the existing closed-loop selection mode in the process of determining the audio coding method, but instead determines the audio coding method to be selected based on the reference linear prediction efficiency of the current audio frame, and the computational complexity of estimating the reference linear prediction efficiency of the current audio frame is usually much smaller than the computational complexity of completely encoding the current audio frame using each audio coding method, compared with the existing mechanism, the above-mentioned technical solution of the embodiment of the present invention is conducive to reducing the computational complexity of audio coding, thereby reducing the overhead of audio coding.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

FIG1 is a schematic flow chart of an audio encoding method provided by one embodiment of the present invention;

FIG2 is a schematic flow chart of another audio encoding method provided by another embodiment of the present invention;

FIG3-a is a schematic structural diagram of an audio encoder provided by an embodiment of the present invention;

FIG3-b is a schematic diagram of the structure of another audio encoder provided by another embodiment of the present invention;

FIG3-c is a schematic structural diagram of another audio encoder provided by another embodiment of the present invention;

FIG3-d is a schematic structural diagram of another audio encoder provided by another embodiment of the present invention;

FIG3-e is a schematic diagram of the structure of another audio encoder provided by another embodiment of the present invention;

FIG3-f is a schematic structural diagram of another audio encoder provided by another embodiment of the present invention;

FIG3-g is a schematic diagram of the structure of another audio encoder provided by another embodiment of the present invention;

FIG3-h is a schematic structural diagram of another audio encoder provided by another embodiment of the present invention;

FIG3-i is a schematic structural diagram of another audio encoder provided by another embodiment of the present invention;

FIG4 is a schematic structural diagram of another audio encoder provided by another embodiment of the present invention;

FIG5 is a schematic structural diagram of another audio encoder provided by another embodiment of the present invention;

FIG6 is a schematic structural diagram of another audio encoder provided by another embodiment of the present invention.

DETAILED DESCRIPTION

The embodiments of the present invention provide an audio encoding method and related devices, in order to reduce the overhead of audio encoding.

In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

The following are detailed descriptions of each.

The terms "first," "second," "third," "fourth," and so on, in the description and claims of the present invention and the accompanying drawings are used to distinguish between different items, not to describe a particular order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements, but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

The following first introduces the audio encoding method provided by an embodiment of the present invention. The executor of the audio encoding method provided by an embodiment of the present invention can be an audio encoder, and the audio encoder can be any device that needs to collect, store or transmit audio signals to the outside, such as a mobile phone, a tablet computer, a personal computer, a laptop computer, etc.

An embodiment of the audio encoding method of the present invention, wherein an audio encoding method may include: estimating a reference linear prediction efficiency of a current audio frame; determining an audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame; and performing audio encoding on the above-mentioned current audio frame according to the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame.

First, please refer to Figure 1, which is a flow chart of an audio encoding method provided by an embodiment of the present invention. As shown in Figure 1, an audio encoding method provided by an embodiment of the present invention may include the following contents:

101. Estimate a reference linear prediction efficiency of the current audio frame.

In practical applications, a variety of available algorithms can be used to estimate the reference linear prediction efficiency of the current audio frame.

In various embodiments of the present invention, a reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. A linear prediction result of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) refers to a linear prediction value of the audio frame. A higher reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree to which the audio frame can be linearly predicted.

In some embodiments of the present invention, the above-mentioned reference linear prediction efficiency includes at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the above-mentioned reference comprehensive linear prediction efficiency is obtained based on the above-mentioned reference long-time linear prediction efficiency and the above-mentioned reference short-time linear prediction efficiency.

The reference long-term linear prediction efficiency of the current audio frame may be obtained based on the long-term linear prediction efficiency of the current audio frame. The reference short-term linear prediction efficiency of the current audio frame may be obtained based on the short-term linear prediction efficiency of the current audio frame. The reference comprehensive linear prediction efficiency of the current audio frame may be obtained based on, for example, the long-term linear prediction efficiency and the short-term linear prediction efficiency of the current audio frame.

It can be understood that the value range of the reference linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x1 (x1 is a positive number). Among them, the value range of the reference long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x2 (x2 is a positive number). The value range of the reference short-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x3 (x3 is a positive number). Among them, the value range of the reference comprehensive linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x4 (x4 is a positive number). Among them, the value range of the long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x5 (x5 is a positive number). The short-term linear prediction efficiency can range from 0 to 1 (i.e., 0% to 100%), or can also range from 0 to x6 (x6 is a positive number). Here, x1, x2, x3, x4, x5, or x6 can be, for example, 0.5, 0.8, 1.5, 2, 5, 10, 50, 100, or other positive numbers. For ease of description, the following examples mainly use the range of 0 to 1 (i.e., 0% to 100%) for each linear prediction efficiency, and other value ranges can be deduced accordingly.

102. Determine an audio coding mode that matches the estimated reference linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, a predetermined mapping relationship may be established between the audio coding mode and the reference linear prediction efficiency of the audio frame. For example, different audio coding modes may correspond to different reference linear prediction efficiencies, or different audio coding modes may correspond to different reference linear prediction efficiency ranges. For example, an audio coding mode that matches the estimated reference linear prediction efficiency of the current audio frame may be determined from at least two audio coding modes.

103. Perform audio encoding on the current audio frame according to an audio encoding method that matches a reference linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, before estimating the reference linear prediction efficiency of the current audio frame, it may be determined whether the current audio frame is a speech audio frame. For example, estimating the reference linear prediction efficiency of the current audio frame may include estimating the reference linear prediction efficiency of the current audio frame when the current audio frame is a non-speech audio frame. Furthermore, estimating the reference linear prediction efficiency of the current audio frame may not distinguish whether the current audio frame is a speech audio frame before estimating the reference linear prediction efficiency of the current audio frame. That is, steps 101 to 103 are performed regardless of whether the current audio frame is a speech audio frame or a non-speech audio frame.

It can be seen that in the technical solution of this embodiment, the reference linear prediction efficiency of the current audio frame is first estimated; the audio coding method that matches it is determined by the estimated reference linear prediction efficiency of the above-mentioned current audio frame, and the audio encoding of the above-mentioned current audio frame is performed according to the determined audio coding method that matches it. Since the above-mentioned scheme does not need to perform the operation of fully encoding the current audio frame using each audio coding method required by the existing closed-loop selection mode in the process of determining the audio coding method, but instead determines the audio coding method to be selected based on the reference linear prediction efficiency of the current audio frame, and the computational complexity of estimating the reference linear prediction efficiency of the current audio frame is usually much smaller than the computational complexity of fully encoding the current audio frame using each audio coding method, compared with the existing mechanism, the above-mentioned scheme of the embodiment of the present invention is conducive to reducing the computational complexity of audio coding, thereby reducing the overhead of audio coding.

In some embodiments of the present invention, a reference comprehensive linear prediction efficiency for an audio frame (e.g., a current audio frame or another audio frame) is obtained based on a reference long-term linear prediction efficiency and a reference short-term linear prediction efficiency for the audio frame. For example, the reference comprehensive linear prediction efficiency for the current audio frame may be the sum, weighted sum (wherein the weight corresponding to the weighted sum may be set as needed, such as 0.5, 1.0, 2.0, 3.0, 5.0, 10.0, or other values), or average value of the reference long-term linear prediction efficiency and the reference short-term linear prediction efficiency for the current audio frame. Of course, the reference comprehensive linear prediction efficiency for the current audio frame may also be obtained using other algorithms based on the reference long-term linear prediction efficiency and the reference short-term linear prediction efficiency for the current audio frame.

In some embodiments of the present invention, audio coding methods based on linear prediction may include Algebraic Code Excited Linear Prediction (ACELP) coding, Transform Coded Excitation (TCX) coding, etc. Audio coding methods not based on linear prediction may include Generic Audio Coding (GAC), which may include, for example, Modified Discrete Cosine Transform (MDCT) coding or Discrete Cosine Transform (DCT) coding.

It is understandable that the specific method for determining the audio coding method that matches the reference linear prediction efficiency of the current audio frame may vary depending on the type of linear prediction efficiency included in the reference linear prediction efficiency of the current audio frame. Some possible embodiments are exemplified below.

For example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction; if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to the first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to the second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a third threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: if the reference long-time linear prediction efficiency of the current audio frame is less than a fourth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a third threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction; if the reference long-time linear prediction efficiency of the current audio frame is less than a fourth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes a reference long-term linear prediction efficiency of the current audio frame, then determining an audio coding mode that matches the reference linear prediction efficiency of the current audio frame includes: determining a first linear prediction efficiency interval within which the reference long-term linear prediction efficiency of the current audio frame falls, and determining a first audio coding mode that has a mapping relationship with the first linear prediction efficiency interval based on a mapping relationship between linear prediction efficiency intervals and audio coding modes based on linear prediction, wherein the first audio coding mode is an audio coding mode that matches the reference linear prediction efficiency of the current audio frame, and the first audio coding mode is an audio coding mode based on linear prediction or an audio coding mode not based on linear prediction. Different linear prediction efficiency intervals correspond to different audio coding modes. For example, assuming there are three linear prediction efficiency intervals, namely 0-30% GAC, 30%-70% TCX, and 70%-100%, if the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-30% (i.e., the first linear prediction efficiency interval is the linear prediction efficiency interval of 0-30%), the audio coding mode corresponding to the linear prediction efficiency interval of 0-30% can be determined to be the audio coding mode (e.g., GAC) that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 30%-70% (i.e., the first linear prediction efficiency interval is the linear prediction efficiency interval of 30%-70%), the audio coding mode (e.g., TCX) that corresponds to the linear prediction efficiency interval of 30%-70% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 70% to 100% (i.e., the first linear prediction efficiency interval is the linear prediction efficiency interval of 70% to 100%), the audio coding method corresponding to the linear prediction efficiency interval of 70% to 100% (e.g., ACELP coding) can be determined as the audio coding method that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between the linear prediction efficiency interval and the linear prediction-based audio coding method can be set according to the needs of different application scenarios.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame may include: if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is less than the fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction; if the reference short-time linear prediction efficiency of the current audio frame is less than the fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: determining the second linear prediction efficiency interval in which the reference short-time linear prediction efficiency of the current audio frame falls, and according to the mapping relationship between the linear prediction efficiency interval and the audio encoding method based on linear prediction, determining a second audio encoding method that has a mapping relationship with the second linear prediction efficiency interval or an audio encoding method that is not based on linear prediction, wherein the above-mentioned second audio encoding method is an audio encoding method that matches the reference linear prediction efficiency of the current audio frame, and the above-mentioned second audio encoding method is an audio encoding method based on linear prediction. For example, assuming there are three linear prediction efficiency intervals, which may be 0-40%, 40%-60%, and 60%-100%, respectively, if the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-40% (i.e., the second linear prediction efficiency interval is the linear prediction efficiency interval of 0-40%), then the audio coding mode corresponding to the linear prediction efficiency interval of 0-40% (e.g., GAC) can be determined as the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 40%-60% (i.e., the second linear prediction efficiency interval is the linear prediction efficiency interval of 40%-60%), then the audio coding mode corresponding to the linear prediction efficiency interval of 40%-60% (e.g., TCX) can be determined as the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency range of 60% to 100% (i.e., the second linear prediction efficiency range is the linear prediction efficiency range of 60% to 100%), the audio coding method corresponding to the linear prediction efficiency range of 60% to 100% (e.g., ACELP coding) is determined as the audio coding method that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between the linear prediction efficiency range and the linear prediction-based audio coding method can be set according to the needs of different application scenarios.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference comprehensive linear prediction efficiency of the current audio frame is greater than or equal to a sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference comprehensive linear prediction efficiency of the current audio frame is less than a sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame may include: if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is less than the sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: determining the third linear prediction efficiency interval in which the reference comprehensive linear prediction efficiency of the current audio frame falls, and according to the mapping relationship between the linear prediction efficiency interval and the audio encoding method based on linear prediction, determining a third audio encoding method that has a mapping relationship with the third linear prediction efficiency interval or an audio encoding method that is not based on linear prediction, wherein the third audio encoding method is an audio encoding method that matches the reference linear prediction efficiency of the current audio frame, and the third audio encoding method is an audio encoding method based on linear prediction. For example, assuming there are three linear prediction efficiency intervals, which may be 0-50%, 50%-80%, and 80%-100%, respectively, if the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-50% (i.e., the third linear prediction efficiency interval is the linear prediction efficiency interval of 0-50%), then the audio coding mode corresponding to the linear prediction efficiency interval of 0-50% (e.g., GAC) can be determined as the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 50%-80% (i.e., the third linear prediction efficiency interval is the linear prediction efficiency interval of 50%-80%), then the audio coding mode corresponding to the linear prediction efficiency interval of 50%-80% (e.g., TCX) can be determined as the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency range of 80% to 100% (i.e., the third linear prediction efficiency range is the linear prediction efficiency range of 80% to 100%), the audio coding method corresponding to the linear prediction efficiency range of 80% to 100% (e.g., ACELP coding) is determined to be the audio coding method that matches the reference linear prediction efficiency of the current audio frame. The same logic can be applied to other scenarios. The mapping relationship between the linear prediction efficiency range and the linear prediction-based audio coding method can be set according to the needs of different application scenarios.

It is understood that the specific values of the various thresholds (e.g., the first threshold, the second threshold, the third threshold, the fourth threshold, the fifth threshold, and the sixth threshold) mentioned in the above examples can be set as needed or according to the application environment and scenario. For example, if the reference long-term linear prediction efficiency of the current audio frame ranges from 0 to 1, the first threshold can be 0.2, 0.5, 0.6, 0.8, 0.9, etc.; if the reference short-term linear prediction efficiency of the current audio frame ranges from 0 to 1, the second threshold can be 0.3, 0.3, 0.6, or 0.8, 0.9, etc. The same applies to other scenarios. Furthermore, the values of the various thresholds can be dynamically and adaptively adjusted as needed. For example, if a linear prediction-based audio coding method (such as TCX or ACELP coding) is preferred for encoding audio frames, the corresponding thresholds (e.g., the first threshold, the second threshold, the third threshold, the fourth threshold, the fifth threshold, and the sixth threshold) can be set relatively small. If an audio coding method not based on linear prediction (such as GAC coding) is preferred for encoding audio frames, the corresponding thresholds (such as the first threshold, the second threshold, the third threshold, the fourth threshold, the fifth threshold, and the sixth threshold) may be set relatively large.

It is understandable that the specific estimation methods of different types of linear prediction efficiencies included in the reference linear prediction efficiency of the current audio frame may be different. The following describes some possible embodiments.

For example, in some embodiments of the present invention, the reference long-time linear prediction efficiency of the current audio frame can be estimated as follows: estimating the long-time linear prediction efficiency of the current audio frame, the long-time linear prediction efficiency of the current audio frame being the reference long-time linear prediction efficiency of the current audio frame.

or,

The reference long-time linear prediction efficiency of the current audio frame is estimated as follows: estimating the long-time linear prediction efficiency of the current audio frame; obtaining the linear prediction efficiencies of N1 historical audio frames of the current audio frame; calculating a first statistical value of the linear prediction efficiencies of the N1 historical audio frames and the long-time linear prediction efficiency of the current audio frame, wherein N1 is a positive integer (for example, N1 may be equal to 1, 2, 3 or other values), and the first statistical value is the reference long-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N11 historical audio frames is at least one of the following linear prediction efficiencies: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; the comprehensive linear prediction efficiency of each of the historical audio frames can be obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each of the historical audio frames (for example, the N11 historical audio frames are audio frames F1, F2 and F3, then the linear prediction efficiency of audio frame F1 is the following linear prediction efficiency of audio frame F1 At least one of the following linear prediction efficiency: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency; the linear prediction efficiency of audio frame F2 is at least one of the following linear prediction efficiency of audio frame F2: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency; the linear prediction efficiency of audio frame F3 is at least one of the following linear prediction efficiency of audio frame F3: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency; the comprehensive linear prediction efficiency of audio frame F1 can be obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of the above-mentioned audio frame F1; the comprehensive linear prediction efficiency of audio frame F2 can be obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of the above-mentioned audio frame F2; the comprehensive linear prediction efficiency of audio frame F3 can be obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of the above-mentioned audio frame F3; and so on for scenarios where N11 takes other values); the above-mentioned N11 historical audio frames are a subset of the above-mentioned N1 historical audio frames (the above-mentioned N11 is less than or equal to the above-mentioned N1). The N1 historical audio frames may be any N1 historical audio frames of the current audio frame, or may be the N1 historical audio frames adjacent to the current audio frame in the time domain. The linear prediction efficiency of the remaining historical audio frames in the N1 historical audio frames other than the N11 historical audio frames may be other types of linear prediction efficiency that are different from the linear prediction efficiency of the N11 historical audio frames, which will not be detailed here. The first statistical value of the linear prediction efficiency of the N1 historical audio frames and the long-term linear prediction efficiency of the current audio frame obtained by calculation may be, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the N1 historical audio frames and the long-term linear prediction efficiency of the current audio frame.

Alternatively, the reference long-time linear prediction efficiency of the current audio frame can be estimated, for example, by: estimating the long-time linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiencies of N2 historical audio frames of the current audio frame; calculating the reference linear prediction efficiencies of the N2 historical audio frames and a second statistical value of the long-time linear prediction efficiency of the current audio frame, wherein N2 is a positive integer (for example, N2 can be equal to 1, 2, 3 or other values), the second statistical value is the reference long-time linear prediction efficiency of the current audio frame, and the reference linear prediction efficiency of each of the N21 historical audio frames is at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of each of the historical audio frames, and the N21 historical audio frames are a subset of the N2 historical audio frames (the N21 is less than or equal to the N2). Among them, the above-mentioned N2 historical audio frames can be any N2 historical audio frames of the above-mentioned current audio frame, or can be N2 historical audio frames adjacent to the above-mentioned current audio frame in the time domain. The linear prediction efficiency of the remaining historical audio frames in the above-mentioned N2 historical audio frames except the above-mentioned N21 historical audio frames can be other types of linear prediction efficiency different from the linear prediction efficiency of the above-mentioned N21 historical audio frames, which are not listed in detail here. The second statistical value of the reference linear prediction efficiency of the above-mentioned N2 historical audio frames and the long-term linear prediction efficiency of the above-mentioned current audio frame calculated is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the reference linear prediction efficiency of the above-mentioned N2 historical audio frames and the long-term linear prediction efficiency of the above-mentioned current audio frame.

Alternatively, the reference long-term linear prediction efficiency of the current audio frame can be estimated, for example, by: estimating the long-term linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiencies of N4 historical audio frames of the current audio frame, and obtaining the linear prediction efficiencies of N3 historical audio frames of the current audio frame; calculating a third statistical value of the linear prediction efficiencies of the N3 historical audio frames, the reference linear prediction efficiencies of the N4 historical audio frames, and the long-term linear prediction efficiency of the current audio frame, wherein the N3 and the N4 are positive integers (for example, the N3 and the N4 can be equal to 1, 2, 3 or other values), and the third statistical value is obtained. The statistical value is the reference long-term linear prediction efficiency of the current audio frame. The linear prediction efficiency of each of the N31 historical audio frames is at least one of the following linear prediction efficiencies: long-term linear prediction efficiency, short-term linear prediction efficiency, and comprehensive linear prediction efficiency. The reference linear prediction efficiency of each of the N41 historical audio frames is at least one of the following linear prediction efficiencies: reference long-term linear prediction efficiency, reference short-term linear prediction efficiency, and reference comprehensive linear prediction efficiency. The N31 historical audio frames are a subset of the N3 historical audio frames, and N31 is less than or equal to N3. The N3 historical audio frames can be any N3 historical audio frames of the current audio frame, or can be N3 historical audio frames that are temporally adjacent to the current audio frame. The linear prediction efficiencies of the remaining N3 historical audio frames, excluding the N31 historical audio frames, can be other types of linear prediction efficiencies different from the linear prediction efficiencies of the N31 historical audio frames, which are not detailed here. The above-mentioned N41 historical audio frames are a subset of the above-mentioned N4 historical audio frames, and the above-mentioned N41 is less than or equal to the above-mentioned N4, wherein the above-mentioned N4 historical audio frames can be any N4 historical audio frames of the above-mentioned current audio frame, or can be N4 historical audio frames adjacent to the above-mentioned current audio frame in the time domain. The linear prediction efficiency of the remaining historical audio frames in the above-mentioned N4 historical audio frames except the above-mentioned N41 historical audio frames can be other types of linear prediction efficiencies different from the linear prediction efficiency of the above-mentioned N41 historical audio frames, which are not listed in detail here. The comprehensive linear prediction efficiency of each of the above-mentioned historical audio frames is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each of the above-mentioned historical audio frames, and the reference comprehensive linear prediction efficiency of each of the above-mentioned historical audio frames is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each of the above-mentioned historical audio frames. The intersection of the above-mentioned N3 historical audio frames and the above-mentioned N4 historical audio frames can be an empty set or not an empty set. The third statistical value calculated for the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame.

For example, in some embodiments of the present invention, the reference short-time linear prediction efficiency of the current audio frame is estimated, for example, by estimating the short-time linear prediction efficiency of the current audio frame, wherein the short-time linear prediction efficiency of the current audio frame is the reference short-time linear prediction efficiency of the current audio frame.

or,

The reference short-time linear prediction efficiency of the current audio frame can be estimated as follows: estimating the short-time linear prediction efficiency of the current audio frame; obtaining the linear prediction efficiencies of N5 historical audio frames of the current audio frame; calculating a fourth statistical value of the linear prediction efficiencies of the N5 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein the N5 is a positive integer (for example, N5 can be equal to 1, 2, 3 or other values), and the fourth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N51 historical audio frames is at least one of the following linear prediction efficiencies: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency, and the comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the long-time linear prediction efficiency and the short-time linear prediction efficiency of each of the historical audio frames, and the N51 historical audio frames are a subset of the N5 historical audio frames (the N51 is less than or equal to the N5). The N5 historical audio frames may be any N5 historical audio frames of the current audio frame, or may be the N5 historical audio frames adjacent to the current audio frame in the time domain. The linear prediction efficiency of the remaining historical audio frames in the N5 historical audio frames other than the N51 historical audio frames may be other types of linear prediction efficiency that are different from the linear prediction efficiency of the N51 historical audio frames, which will not be detailed here. The fourth statistical value of the linear prediction efficiency of the N5 historical audio frames and the short-time linear prediction efficiency of the current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the N5 historical audio frames and the short-time linear prediction efficiency of the current audio frame.

or,

The reference short-time linear prediction efficiency of the current audio frame can be estimated in the following manner: estimating the short-time linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiencies of N6 historical audio frames of the current audio frame; calculating a fifth statistical value of the reference linear prediction efficiencies of the N6 historical audio frames and the short-time linear prediction efficiency of the current audio frame, where N6 is a positive integer (for example, N6 can be equal to 1, 2, 3 or other values), and the fifth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each of the N61 historical audio frames is at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of each of the historical audio frames, and the N61 historical audio frames are a subset of the N6 historical audio frames (the N61 is less than or equal to the N6). Among them, the above-mentioned N6 historical audio frames can be any N6 historical audio frames of the above-mentioned current audio frame, or can be N6 historical audio frames adjacent to the above-mentioned current audio frame in the time domain. The linear prediction efficiency of the remaining historical audio frames in the above-mentioned N6 historical audio frames except the above-mentioned N61 historical audio frames can be other types of linear prediction efficiency different from the linear prediction efficiency of the above-mentioned N61 historical audio frames, which are not listed in detail here. Among them, the fifth statistical value of the reference linear prediction efficiency of the above-mentioned N6 historical audio frames and the short-time linear prediction efficiency of the above-mentioned current audio frame calculated can be the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the reference linear prediction efficiency of the above-mentioned N6 historical audio frames and the short-time linear prediction efficiency of the above-mentioned current audio frame.

or,

The reference short-time linear prediction efficiency of the current audio frame can be estimated as follows: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of N8 historical audio frames of the current audio frame; obtain the linear prediction efficiency of N7 historical audio frames of the current audio frame; calculate the linear prediction efficiency of the N7 historical audio frames, the reference linear prediction efficiency of the N8 historical audio frames and the sixth statistical value of the short-time linear prediction efficiency of the current audio frame, where N7 and N8 are positive integers (for example, N7 and N8 can be equal to 1, 2, 3 or other values), the sixth statistical value is the reference short-time linear prediction efficiency of the current audio frame, and the linear prediction efficiency of each historical audio frame in the N71 historical audio frames is the following linear prediction efficiency. At least one of the following efficiencies: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency; the reference linear prediction efficiency of each of the N81 historical audio frames is at least one of the following linear prediction efficiencies: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency, and reference comprehensive linear prediction efficiency; the comprehensive linear prediction efficiency of each of the above historical audio frames is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each of the above historical audio frames; wherein, the reference comprehensive linear prediction efficiency of each of the above historical audio frames is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each of the above historical audio frames; the above N71 historical audio frames are a subset of the above N7 historical audio frames (the above N71 is less than or equal to the above N7); wherein, the above N7 historical audio frames can be any N7 historical audio frames of the above current audio frame, or can be N7 historical audio frames adjacent to the above current audio frame in the time domain. The linear prediction efficiency of the remaining historical audio frames in the above-mentioned N7 historical audio frames except the above-mentioned N71 historical audio frames may be other types of linear prediction efficiency different from the linear prediction efficiency of the above-mentioned N71 historical audio frames, which will not be listed in detail here. The above-mentioned N81 historical audio frames are a subset of the above-mentioned N8 historical audio frames (the above-mentioned N81 is less than or equal to the above-mentioned N8). Among them, the above-mentioned N8 historical audio frames may be any N8 historical audio frames of the above-mentioned current audio frame, or may be N8 historical audio frames adjacent to the above-mentioned current audio frame in the time domain. The linear prediction efficiency of the remaining historical audio frames in the above-mentioned N8 historical audio frames except the above-mentioned N81 historical audio frames may be other types of linear prediction efficiency different from the linear prediction efficiency of the above-mentioned N81 historical audio frames, which will not be listed in detail here. The intersection of the above-mentioned N7 historical audio frames and the above-mentioned N8 historical audio frames may be an empty set or not an empty set. Among them, the sixth statistical value of the calculated linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

In some embodiments of the present invention, the linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. The linear prediction result of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) refers to the linear prediction value of the audio frame. A higher linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree of linear prediction for the audio frame.

In some embodiments of the present invention, the estimating and obtaining the short-time linear prediction efficiency of the current audio frame may include: obtaining the short-time linear prediction efficiency of the current audio frame based on a linear prediction residual of the current audio frame.

In some embodiments of the present invention, obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame may, for example, include calculating an energy change rate before and after performing short-time linear prediction on the current audio frame, wherein the calculated energy change rate is the short-time linear prediction efficiency of the current audio frame. Alternatively, the short-time linear prediction efficiency of the current audio frame is obtained by transforming the calculated energy change rate, wherein the energy of the current audio frame after performing short-time linear prediction is the energy of the linear prediction residual of the current audio frame. For example, a mapping relationship may exist between the energy change rate and the short-time linear prediction efficiency of the current audio frame. Based on the mapping relationship between the energy change rate and the short-time linear prediction efficiency of the current audio frame, the short-time linear prediction efficiency of the current audio frame having a mapping relationship with the calculated energy change rate may be obtained. Generally speaking, a greater energy change rate before and after performing short-time linear prediction on the current audio frame indicates a higher short-time linear prediction efficiency of the current audio frame.

For example, the energy change rate of the current audio frame before and after short-time linear prediction may be a ratio of the energy of the current audio frame before short-time linear prediction to the energy of the linear prediction residual of the current audio frame, or the inverse of the ratio. Generally speaking, a larger ratio of the energy of the current audio frame before short-time linear prediction divided by the energy of the linear prediction residual of the current audio frame indicates a higher short-time linear prediction efficiency for the current audio frame.

In some embodiments of the present invention, the above-mentioned estimation of the long-time linear prediction efficiency of the current audio frame may include: obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal, the above-mentioned correlation is the long-time linear prediction efficiency of the current audio frame, or the long-time linear prediction efficiency of the current audio frame is obtained based on the above-mentioned transformation. In which, the above-mentioned first historical linear prediction signal is the first historical linear prediction excitation or the first historical linear prediction residual; the above-mentioned first historical linear prediction residual is the linear prediction residual of the historical audio frame of the above-mentioned current audio frame (for example, the above-mentioned first historical linear prediction residual can be the linear prediction residual of a historical audio frame having the same or similar duration as the above-mentioned current audio frame and being the current audio frame, or the above-mentioned first historical linear prediction residual can be the linear prediction residual of a part of the continuous audio signal of two adjacent historical audio frames of the above-mentioned current audio frame, and being the current audio frame), and the above-mentioned first historical linear prediction excitation is the linear prediction excitation of the historical audio frame of the above-mentioned current audio frame (for example, the above-mentioned first historical linear prediction excitation can be the linear prediction excitation of a historical audio frame having the same or similar duration as the above-mentioned current audio frame and being the current audio frame, or the above-mentioned first historical linear prediction excitation can be the linear prediction excitation of a part of the continuous audio signal of two adjacent historical audio frames of the current audio frame). For example, there is a mapping relationship between the correlation and the long-term linear prediction efficiency of the audio frame. Based on the mapping relationship between the correlation and the long-term linear prediction efficiency of the audio frame, the long-term linear prediction efficiency of the current audio frame having a mapping relationship with the calculated correlation can be obtained.

There are various ways to obtain the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal.

For example, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal may include: calculating the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

Alternatively, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame may include: multiplying the linear prediction residual of the current audio frame by a gain factor to obtain the gain linear prediction residual of the current audio frame, and calculating the correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal, wherein the calculated correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

Alternatively, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal may include: multiplying the first historical linear prediction signal by a gain factor to obtain the gained first historical linear prediction signal, and calculating the correlation between the linear prediction residual of the current audio frame and the gained first historical linear prediction signal, wherein the calculated correlation between the linear prediction residual of the current audio frame and the gained first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

The first historical linear prediction excitation or the first historical linear prediction residual may be determined based on the pitch of the current audio frame. For example, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction excitations and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction excitation and the linear prediction residual of the current audio frame. For example, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction residuals and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction residual and the linear prediction residual of the current audio frame.

Generally speaking, a greater correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal indicates a higher long-term linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the above-mentioned correlation is, for example, a cross-correlation function value in the time domain and/or a cross-correlation function value in the frequency domain, or the above-mentioned correlation may be a distortion in the time domain and/or a distortion in the frequency domain (wherein the distortion in the frequency domain may also be referred to as spectral distortion).

In some embodiments of the present invention, the above-mentioned distortion in the frequency domain may be the sum or weighted sum of the distortions of K1 frequency points in the frequency domain, or the above-mentioned distortion in the frequency domain may be the sum or weighted sum of the distortions on K2 sub-bands in the frequency domain, and the above-mentioned K1 and the above-mentioned K2 are positive integers.

Generally speaking, the larger the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the time domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the larger the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the frequency domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the smaller the distortion between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the frequency domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the smaller the distortion between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the time domain, the higher the long-term linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the weighting coefficient corresponding to the weighted sum of the distortion values is a perceptual weighting coefficient reflecting a psychoacoustic model. Of course, the weighting coefficient corresponding to the weighted sum of the distortion values can also be other weighting coefficients set based on actual needs. Testing has found that using a perceptual weighting coefficient helps make the calculated distortion more consistent with subjective quality, thereby improving performance.

In some embodiments of the present invention, the first historical linear prediction excitation may be a linear prediction excitation generated by audio encoding a historical audio frame of the current audio frame using a linear prediction-based encoding method.

In some embodiments of the present invention, the first historical linear prediction residual may be obtained based on a time domain signal of a first historical audio frame of the current audio frame and linear prediction coefficients of the first historical audio frame, wherein the linear prediction coding coefficients of the first historical audio frame are quantized linear prediction coefficients or unquantized linear prediction coefficients. Since the quantized linear prediction coefficients generally contribute to the final quality in actual encoding and decoding processes, using the quantized linear prediction coefficients to calculate the linear prediction residual facilitates more accurate calculated correlation.

In some embodiments of the present invention, the linear prediction residual of the current audio frame may be obtained based on the time domain signal of the current audio frame and the linear prediction coefficient of the current audio frame, wherein the linear prediction coefficient of the current audio frame may be a quantized linear prediction coefficient or an unquantized linear prediction coefficient. Since the quantized linear prediction coefficients generally contribute to the final quality in actual encoding and decoding processes, using the quantized linear prediction coefficients to calculate the linear prediction residual facilitates more accurate calculated correlation.

In some embodiments of the present invention, the first historical linear prediction excitation may be a superposition of an adaptive codebook excitation and a fixed codebook excitation, or the first historical linear prediction excitation may be an adaptive codebook excitation. Alternatively, the first historical linear prediction excitation may be another type of codebook excitation.

It is understandable that in each embodiment of the present invention, the historical audio frame of an audio frame (such as the current audio frame or an audio frame that is located before or after the current audio frame in the time domain) refers to the audio frame that is located before the audio frame in the time domain in the same audio stream. It can be seen that the historical audio frame is a relative concept. For example, assuming that the order of the four audio frames contained in the same audio stream in the time domain is audio frame y1->audio frame y2->audio frame y3->audio frame y4, then audio frame y1, audio frame y2, and audio frame y3 are all historical audio frames of audio frame y4, audio frame y1 and audio frame y2 are both historical audio frames of audio frame y3, and audio frame y1 is the historical audio frame of audio frame y2. It is understandable that audio frame y4 is not the historical audio frame of audio frame y3, nor is audio frame y4 the historical audio frame of audio frame y2 and audio frame y1, and other scenarios can be deduced by analogy.

To facilitate a better understanding of the above technical aspects of the embodiments of the present invention, some specific application scenarios are given as examples below.

First, please refer to Figure 2, which is a flowchart of an audio encoding method provided by an embodiment of the present invention. As shown in Figure 2, an audio encoding method provided by an embodiment of the present invention may include the following contents:

201. Determine whether the current audio frame is a voice audio frame.

If so, execute step 202.

If not, execute step 203.

202. Perform audio encoding on the current audio frame based on a speech encoding method.

In some embodiments of the present invention, if the current audio frame is a speech audio frame, audio encoding may be performed on the current audio frame based on Algebraic Code Excited Linear Prediction (ACELP) encoding. For example, if the current audio frame is a speech audio frame, the current audio frame may be input into an ACELP subcoder for audio encoding. The ACELP subcoder is a subcoder that uses ACELP encoding.

203. Estimate a reference linear prediction efficiency of the current audio frame.

Therein, a variety of algorithms can be used to estimate the reference linear prediction efficiency of the current audio frame.

In various embodiments of the present invention, a reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. A linear prediction result of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) refers to a linear prediction value of the audio frame. A higher reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree to which the audio frame can be linearly predicted.

In some embodiments of the present invention, the above-mentioned reference linear prediction efficiency includes at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the above-mentioned reference comprehensive linear prediction efficiency is obtained based on the above-mentioned reference long-time linear prediction efficiency and the above-mentioned reference short-time linear prediction efficiency.

The reference long-term linear prediction efficiency of the current audio frame may be obtained based on the long-term linear prediction efficiency of the current audio frame. The reference short-term linear prediction efficiency of the current audio frame may be obtained based on the short-term linear prediction efficiency of the current audio frame. The reference comprehensive linear prediction efficiency of the current audio frame may be obtained based on, for example, the long-term linear prediction efficiency and the short-term linear prediction efficiency of the current audio frame.

It can be understood that the value range of the reference linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x1 (x1 is a positive number). Among them, the value range of the reference long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x2 (x2 is a positive number). The value range of the reference short-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x3 (x3 is a positive number). Among them, the value range of the reference comprehensive linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x4 (x4 is a positive number). Among them, the value range of the long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x5 (x5 is a positive number). The short-time linear prediction efficiency may range from 0 to 1 (i.e., 0% to 100%), or may range from 0 to x6 (x6 is a positive number). Here, x1, x2, x3, x4, x5, or x6 may be, for example, 0.5, 0.8, 1.5, 2, 5, 10, 50, 100, or other positive numbers.

204. Determine an audio coding mode that matches the estimated reference linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, a mapping relationship may be established between audio coding modes and reference linear prediction efficiencies of audio frames. For example, different audio coding modes may correspond to different reference linear prediction efficiencies. For example, an audio coding mode that matches the estimated reference linear prediction efficiency of the current audio frame may be determined from at least two audio coding modes.

The audio coding method that matches the estimated reference linear prediction efficiency of the current audio frame may be Transform Coded Excitation (TCX) or Generic Audio Coding (GAC). GAC may be, for example, Modified Discrete Cosine Transform (MDCT) coding.

205. Perform audio encoding on the current audio frame according to the determined audio encoding method.

As can be seen, in the technical solution of this embodiment, it is first determined whether the current audio frame is a speech audio frame. If the current audio frame is a speech audio frame, the current audio frame is audio-encoded based on the speech coding mode. If the current audio frame is a non-speech audio frame, the reference linear prediction efficiency of the current audio frame is first estimated; the audio coding mode that matches the current audio frame is determined based on the estimated reference linear prediction efficiency of the current audio frame, and the current audio frame is audio-encoded according to the determined matching audio coding mode. Since the above scheme does not need to perform the operation of fully encoding the current audio frame using each audio coding mode required by the existing closed-loop selection mode during the process of determining the audio coding mode, but instead determines the audio coding mode to be selected based on the reference linear prediction efficiency of the current audio frame, and the computational complexity of estimating the reference linear prediction efficiency of the current audio frame is generally much less than the computational complexity of fully encoding the current audio frame using each audio coding mode, compared to the existing mechanism, the above scheme of the embodiment of the present invention is conducive to reducing the computational complexity of audio coding, thereby reducing the audio coding overhead.

In some embodiments of the present invention, the reference comprehensive linear prediction efficiency of the current audio frame may be, for example, the sum of the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, a weighted sum (wherein, the weight corresponding to the weighted sum here may be set according to actual needs, where one weight may be, for example, 0.5, 1., 2, 3, 5, 10 or other values) or an average value.

It is understandable that the specific method for determining the audio coding method that matches the reference linear prediction efficiency of the current audio frame may vary depending on the type of linear prediction efficiency included in the reference linear prediction efficiency of the current audio frame. Some possible embodiments are exemplified below.

For example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction; if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to the first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to the second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a third threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: if the reference long-time linear prediction efficiency of the current audio frame is less than a fourth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a third threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction; if the reference long-time linear prediction efficiency of the current audio frame is less than a fourth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes a reference long-term linear prediction efficiency of the current audio frame, then determining an audio coding mode that matches the reference linear prediction efficiency of the current audio frame includes: determining a first linear prediction efficiency interval within which the reference long-term linear prediction efficiency of the current audio frame falls, and determining a first audio coding mode that has a mapping relationship with the first linear prediction efficiency interval based on a mapping relationship between linear prediction efficiency intervals and audio coding modes based on linear prediction, wherein the first audio coding mode is an audio coding mode that matches the reference linear prediction efficiency of the current audio frame, and the first audio coding mode is an audio coding mode based on linear prediction or an audio coding mode not based on linear prediction. Different linear prediction efficiency intervals correspond to different audio coding modes. For example, assuming there are three linear prediction efficiency intervals, which can be 0-30%, 30%-70%, and 70%-100%, respectively, if the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-30% (i.e., the first linear prediction efficiency interval is the linear prediction efficiency interval of 0-30%), the audio coding mode corresponding to the linear prediction efficiency interval of 0-30% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 30%-70% (i.e., the first linear prediction efficiency interval is the linear prediction efficiency interval of 30%-70%), the audio coding mode corresponding to the linear prediction efficiency interval of 30%-70% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between linear prediction efficiency intervals and linear prediction-based audio coding modes can be set according to the needs of different application scenarios.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame may include: if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is less than the fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction; if the reference short-time linear prediction efficiency of the current audio frame is less than the fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: determining the second linear prediction efficiency interval in which the reference short-time linear prediction efficiency of the current audio frame falls, and determining a second audio encoding method having a mapping relationship with the second linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio encoding method based on linear prediction, wherein the second audio encoding method is an audio encoding method that matches the reference linear prediction efficiency of the current audio frame, and the second audio encoding method is an audio encoding method based on linear prediction or an audio encoding method not based on linear prediction. For example, assuming there are three linear prediction efficiency intervals, which can be 0-40%, 40%-60%, and 60%-100%, respectively. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-40% (i.e., the second linear prediction efficiency interval is the linear prediction efficiency interval of 0-40%), the audio coding mode corresponding to the linear prediction efficiency interval of 0-40% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 40%-60% (i.e., the second linear prediction efficiency interval is the linear prediction efficiency interval of 40%-60%), the audio coding mode corresponding to the linear prediction efficiency interval of 40%-60% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between linear prediction efficiency intervals and linear prediction-based audio coding modes can be set according to the needs of different application scenarios.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference comprehensive linear prediction efficiency of the current audio frame is greater than or equal to a sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may include: if the reference comprehensive linear prediction efficiency of the current audio frame is less than a sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame may include: if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is less than the sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame includes: determining the third linear prediction efficiency interval in which the reference comprehensive linear prediction efficiency of the current audio frame falls, and determining a third audio encoding method that has a mapping relationship with the third linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio encoding method based on linear prediction, wherein the third audio encoding method is an audio encoding method that matches the reference linear prediction efficiency of the current audio frame, and the third audio encoding method is an audio encoding method based on linear prediction or an audio encoding method not based on linear prediction. For example, assuming there are three linear prediction efficiency intervals, which can be 0-50%, 50%-80%, and 80%-100%, respectively. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-50% (i.e., the third linear prediction efficiency interval is the linear prediction efficiency interval of 0-50%), the audio coding mode corresponding to the linear prediction efficiency interval of 0-50% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 50%-80% (i.e., the third linear prediction efficiency interval is the linear prediction efficiency interval of 50%-80%), the audio coding mode corresponding to the linear prediction efficiency interval of 50%-80% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between linear prediction efficiency intervals and linear prediction-based audio coding modes can be set according to the needs of different application scenarios.

It is understandable that the specific estimation methods of different types of linear prediction efficiencies included in the reference linear prediction efficiency of the current audio frame may be different. The following describes some possible embodiments.

For example, in some embodiments of the present invention, the reference long-time linear prediction efficiency of the current audio frame can be estimated as follows: estimating the long-time linear prediction efficiency of the current audio frame, the long-time linear prediction efficiency of the current audio frame being the reference long-time linear prediction efficiency of the current audio frame.

or,

The reference long-time linear prediction efficiency of the current audio frame is estimated in the following manner: estimating the long-time linear prediction efficiency of the current audio frame; obtaining the linear prediction efficiencies of N1 historical audio frames of the current audio frame; calculating the linear prediction efficiencies of the N1 historical audio frames and the first statistical value of the long-time linear prediction efficiency of the current audio frame, wherein the N1 is a positive integer, and the first statistical value is the reference long-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N11 historical audio frames is at least one of the following linear prediction efficiencies of each of the historical audio frames: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; the comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each of the historical audio frames, and the N11 historical audio frames are a subset of the N1 historical audio frames. Among them, the first statistical value of the calculated linear prediction efficiency of the above N1 historical audio frames and the long-time linear prediction efficiency of the above current audio frame can be, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the above N1 historical audio frames and the long-time linear prediction efficiency of the above current audio frame.

Alternatively, the reference long-time linear prediction efficiency of the current audio frame can be estimated, for example, in the following manner: estimating the long-time linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiencies of N2 historical audio frames of the current audio frame; calculating the reference linear prediction efficiencies of the N2 historical audio frames and the second statistical value of the long-time linear prediction efficiency of the current audio frame, wherein the N2 is a positive integer, and the second statistical value is the reference long-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each of the N21 historical audio frames is at least one of the following linear prediction efficiencies of each of the historical audio frames: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of each of the historical audio frames, and the N21 historical audio frames are a subset of the N2 historical audio frames. The second statistical value of the reference linear prediction efficiency of the above N2 historical audio frames and the long-time linear prediction efficiency of the above current audio frame is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the reference linear prediction efficiency of the above N2 historical audio frames and the long-time linear prediction efficiency of the above current audio frame.

Alternatively, the reference long-time linear prediction efficiency of the current audio frame can be estimated, for example, by: estimating the long-time linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiencies of N4 historical audio frames of the current audio frame, and obtaining the linear prediction efficiencies of N3 historical audio frames of the current audio frame; calculating a third statistical value of the linear prediction efficiencies of the N3 historical audio frames, the reference linear prediction efficiencies of the N4 historical audio frames, and the long-time linear prediction efficiency of the current audio frame, wherein the N3 and the N4 are positive integers, the third statistical value is the reference long-time linear prediction efficiency of the current audio frame, and the linear prediction efficiency of each of the N31 historical audio frames is at least one of the following linear prediction efficiencies of each of the historical audio frames: long-time linear prediction Efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; wherein, the reference linear prediction efficiency of each historical audio frame in the N41 historical audio frames is at least one of the following linear prediction efficiencies of each historical audio frame: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the N31 historical audio frames are a subset of the N3 historical audio frames, the N41 historical audio frames are a subset of the N4 historical audio frames, the comprehensive linear prediction efficiency of each historical audio frame is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each historical audio frame, and the reference comprehensive linear prediction efficiency of each historical audio frame is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each historical audio frame. The intersection of the N3 historical audio frames and the N4 historical audio frames may be an empty set or not an empty set. The third statistical value calculated for the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame.

For example, in some embodiments of the present invention, the reference short-time linear prediction efficiency of the current audio frame is estimated, for example, by estimating the short-time linear prediction efficiency of the current audio frame, wherein the short-time linear prediction efficiency of the current audio frame is the reference short-time linear prediction efficiency of the current audio frame.

or,

The reference short-time linear prediction efficiency of the current audio frame can be estimated in the following manner: estimating the short-time linear prediction efficiency of the current audio frame; obtaining the linear prediction efficiencies of N5 historical audio frames of the current audio frame; calculating the fourth statistical value of the linear prediction efficiencies of the N5 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein the N5 is a positive integer, and the fourth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N51 historical audio frames is at least one of the following linear prediction efficiencies of each of the historical audio frames: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency, and the comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each of the historical audio frames, and the N51 historical audio frames are a subset of the N5 historical audio frames. Among them, the fourth statistical value of the calculated linear prediction efficiency of the above N5 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the linear prediction efficiency of the above N5 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

or,

The reference short-time linear prediction efficiency of the current audio frame can be estimated in the following manner: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of N6 historical audio frames of the current audio frame; calculate the fifth statistical value of the reference linear prediction efficiencies of the N6 historical audio frames and the short-time linear prediction efficiency of the current audio frame, where N6 is a positive integer and the fifth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each of the N61 historical audio frames is at least one of the following linear prediction efficiencies of each of the historical audio frames: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of each of the historical audio frames, and the N61 historical audio frames are a subset of the N6 historical audio frames. Among them, the fifth statistical value of the reference linear prediction efficiency of the above N6 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted average of the reference linear prediction efficiency of the above N6 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

or,

The reference short-time linear prediction efficiency of the current audio frame can be estimated in the following manner: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of N8 historical audio frames of the current audio frame; obtain the linear prediction efficiency of N7 historical audio frames of the current audio frame; calculate the sixth statistical value of the linear prediction efficiency of the N7 historical audio frames, the reference linear prediction efficiency of the N8 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein N7 and N8 are positive integers, the sixth statistical value is the reference short-time linear prediction efficiency of the current audio frame, and the linear prediction efficiency of each historical audio frame in the N71 historical audio frames is at least one of the following linear prediction efficiencies of each historical audio frame: long-time linear prediction efficiency, Short-time linear prediction efficiency and comprehensive linear prediction efficiency. The reference linear prediction efficiency of each of the N81 historical audio frames is at least one of the following linear prediction efficiencies for each of the above-mentioned historical audio frames: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency, and reference comprehensive linear prediction efficiency. The comprehensive linear prediction efficiency of each of the above-mentioned historical audio frames is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each of the above-mentioned historical audio frames, wherein the reference comprehensive linear prediction efficiency of each of the above-mentioned historical audio frames is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each of the above-mentioned historical audio frames. The above-mentioned N71 historical audio frames are a subset of the above-mentioned N7 historical audio frames, and the above-mentioned N81 historical audio frames are a subset of the above-mentioned N8 historical audio frames. The intersection of the above-mentioned N7 historical audio frames and the above-mentioned N8 historical audio frames may be an empty set or a non-empty set. Among them, the sixth statistical value of the calculated linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

In some embodiments of the present invention, the linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. The linear prediction result of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) refers to the linear prediction value of the audio frame. A higher linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree of linear prediction for the audio frame.

In some embodiments of the present invention, the estimating and obtaining the short-time linear prediction efficiency of the current audio frame may include: obtaining the short-time linear prediction efficiency of the current audio frame based on a linear prediction residual of the current audio frame.

In some embodiments of the present invention, obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame may, for example, include calculating an energy change rate before and after performing short-time linear prediction on the current audio frame, wherein the calculated energy change rate is the short-time linear prediction efficiency of the current audio frame. Alternatively, the short-time linear prediction efficiency of the current audio frame is obtained by transforming the calculated energy change rate, wherein the energy of the current audio frame after performing short-time linear prediction is the energy of the linear prediction residual of the current audio frame. For example, a mapping relationship may exist between the energy change rate and the short-time linear prediction efficiency of the current audio frame. Based on the mapping relationship between the energy change rate and the short-time linear prediction efficiency of the current audio frame, the short-time linear prediction efficiency of the current audio frame having a mapping relationship with the calculated energy change rate may be obtained. Generally speaking, a greater energy change rate before and after performing short-time linear prediction on the current audio frame indicates a higher short-time linear prediction efficiency of the current audio frame.

For example, the energy change rate of the current audio frame before and after short-time linear prediction may be a ratio of the energy of the current audio frame before short-time linear prediction to the energy of the linear prediction residual of the current audio frame, or the inverse of the ratio. Generally speaking, a larger ratio of the energy of the current audio frame before short-time linear prediction divided by the energy of the linear prediction residual of the current audio frame indicates a higher short-time linear prediction efficiency for the current audio frame.

In some embodiments of the present invention, estimating the long-term linear prediction efficiency of the current audio frame may include: obtaining a correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal, wherein the correlation is the long-term linear prediction efficiency of the current audio frame, or the long-term linear prediction efficiency of the current audio frame is obtained based on the transformation. The first historical linear prediction signal is a first historical linear prediction excitation or a first historical linear prediction residual; the first historical linear prediction residual is a linear prediction residual of historical audio frames of the current audio frame; and the first historical linear prediction excitation is a linear prediction excitation of historical audio frames of the current audio frame. For example, if there is a mapping relationship between the correlation and the long-term linear prediction efficiency of the audio frame, based on the mapping relationship between the correlation and the long-term linear prediction efficiency of the audio frame, the long-term linear prediction efficiency of the current audio frame having a mapping relationship with the calculated correlation can be obtained.

In some embodiments of the present invention, an analysis filter A(Z) may be used to filter the time domain signal of the current audio frame to obtain a linear prediction residual R of the current audio frame, where the filter coefficients of the filter A(Z) are the linear prediction coefficients of the current audio frame.

The specific formula is as follows:

Wherein, S(i) in Formula 1 represents the signal of the i-th time domain sample point of the current audio frame, a(k) represents the k-th order linear prediction coefficient of the current audio frame, M is the total order of the filter, the above N is the time domain length of the current audio frame, and R(i) represents the linear prediction residual of the i-th time domain sample point of the current audio frame.

It can be understood that the linear prediction residual of any audio frame (such as the current audio frame or the historical audio frame of the current audio frame) can be obtained through the above example method.

For example, the linear prediction excitation or linear prediction residual of each audio frame or part of the audio frame can be cached so as to serve as the historical linear prediction excitation or historical linear prediction residual that may be used in the possible next audio frame to calculate its correlation with the linear prediction residual of the next audio frame.

There are various ways to obtain the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal.

For example, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal may include: calculating the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

Alternatively, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame may include: multiplying the linear prediction residual of the current audio frame by a gain factor to obtain the gain linear prediction residual of the current audio frame, and calculating the correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal, wherein the calculated correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

Alternatively, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal may include: multiplying the first historical linear prediction signal by a gain factor to obtain the gained first historical linear prediction signal, and calculating the correlation between the linear prediction residual of the current audio frame and the gained first historical linear prediction signal, wherein the calculated correlation between the linear prediction residual of the current audio frame and the gained first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

The first historical linear prediction excitation or the first historical linear prediction residual may be determined based on the pitch of the current audio frame. For example, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction excitations and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction excitation and the linear prediction residual of the current audio frame. For example, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction residuals and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction residual and the linear prediction residual of the current audio frame.

In some embodiments of the present invention, the above correlation is, for example, a cross-correlation function value in the time domain and/or a cross-correlation function value in the frequency domain.

In an optional embodiment of the present invention, when calculating the cross-correlation function value in the frequency domain, a time-frequency transform (such as a discrete Fourier transform (DFT) or a discrete cosine transform (DCT)) may be performed on the linear prediction residual of the current audio frame to obtain a frequency domain signal of the linear prediction residual of the current audio frame, and a time-frequency transform (such as a DFT or DCT) may be performed on the first historical linear prediction signal to obtain a frequency domain signal of the first historical linear prediction signal. An example correlation calculation formula is provided below, as shown in Formula 1:

Wherein, in the above formula 2, C represents the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the time domain, R(i) represents the linear prediction residual of the i-th time domain sample of the current audio frame, E(i) represents the signal of the i-th time domain sample of the first historical linear prediction signal, and N represents the total number of time domain samples of an audio frame. Alternatively, in the above formula 2, C represents the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the frequency domain, R(i) represents the i-th spectral envelope of the linear prediction residual of the current audio frame, E(i) represents the signal of the i-th spectral envelope of the first historical linear prediction signal, and N represents the total number of spectral envelopes of an audio frame. Of course, the present invention does not limit other correlation calculation methods.

In another embodiment of the present invention in which correlation is calculated in the frequency domain, in order to better overcome pitch jitter, one of the signals R(i) or E(i) may be shifted before calculating the cross-correlation, as shown in the following formula 3:

Based on Formula 2, Formula 3 further performs a shift process on E(i). j represents the shift amount and can be an integer. The shift process on R(i) is similar.

In other embodiments of the present invention, the above correlation may be, for example, distortion in the time domain and/or distortion in the frequency domain.

In an optional embodiment of the present invention, when calculating the frequency domain distortion, a time-frequency transform (e.g., DFT or DCT) may be performed on the linear prediction residual of the current audio frame to obtain a frequency domain signal of the linear prediction residual of the current audio frame, and a time-frequency transform (e.g., DFT or DCT) may be performed on the first historical linear prediction signal to obtain a frequency domain signal of the first historical linear prediction signal. A distortion D is calculated between the frequency domain signal of the linear prediction residual of the current audio frame and the frequency domain signal of the first historical linear prediction signal.

The smaller the distortion D, the stronger the correlation and the higher the efficiency of long-term linear prediction. The following example shows a calculation formula for distortion D, as shown in Formula 4:

In Formula 4, N may represent the total number of time-domain samples in an audio frame, R(k) represents the linear prediction residual of the k-th time-domain sample of the current audio frame, and E(k) represents the signal of the k-th time-domain sample of the first historical linear prediction signal. Alternatively, N in Formula 4 may represent the total number of spectral envelopes in an audio frame, R(k) represents the k-th spectral envelope of the linear prediction residual of the current audio frame, and E(k) represents the k-th spectral envelope of the first historical linear prediction signal.

The following are two other distortion D calculation formulas, as shown in Formula 5 or Formula 6:

In Formulas 5 and 6, N may represent the total number of time-domain samples in an audio frame, R(k) represents the linear prediction residual of the k-th time-domain sample of the current audio frame, and E(k) represents the signal of the k-th time-domain sample of the first historical linear prediction signal. Alternatively, N in Formulas 5 and 6 may represent the total number of spectral envelopes in an audio frame, R(k) represents the k-th spectral envelope of the linear prediction residual of the current audio frame, and E(k) represents the k-th spectral envelope of the first historical linear prediction signal.

In Formulas 5 and 6, G represents a gain factor. By selecting an appropriate value for G, the distortion D can be minimized. In Formula 4, the gain factor G is applied to E(k), and in Formula 5, the gain factor G is applied to R(k).

In some embodiments of the present invention, the above-mentioned distortion in the frequency domain may be the sum or weighted sum of the distortions of K1 frequency points in the frequency domain, or the above-mentioned distortion in the frequency domain may be the sum or weighted sum of the distortions on K2 sub-bands in the frequency domain, and the above-mentioned K1 and the above-mentioned K2 are positive integers.

The following are three examples of calculation formulas for distortion D, as shown in Formula 7, Formula 8, or Formula 9:

In Formula 7 to Formula 9, P(k) is a set of weighting coefficients. P(k) may be a set of perceptual weighting coefficients or other weighting coefficients reflecting a psychoacoustic model.

In Formulas 7 to 9, N, R(k), E(k), and G have the same meanings as in Formula 5.

In some embodiments of the present invention, the first historical linear prediction excitation may be a linear prediction excitation generated by audio encoding the historical audio frame s of the current audio frame using a linear prediction-based encoding method.

In some embodiments of the present invention, the first historical linear prediction residual can be obtained based on the time domain signal of the first historical audio frame of the current audio frame and the linear prediction coefficient of the first historical audio frame, wherein the linear prediction coding coefficient of the first historical audio frame is a quantized linear prediction coefficient or an unquantized linear prediction coefficient.

In some embodiments of the present invention, the linear prediction residual of the current audio frame may be obtained based on the time domain signal of the current audio frame and the linear prediction coefficient of the current audio frame, wherein the linear prediction coefficient of the current audio frame may be a quantized linear prediction coefficient or an unquantized linear prediction coefficient.

In some embodiments of the present invention, the first historical linear prediction excitation may be a superposition of an adaptive codebook excitation and a fixed codebook excitation, or the first historical linear prediction excitation may be an adaptive codebook excitation.

Related devices for implementing the above solutions are also provided below.

Refer to FIG3 - a , which is a schematic structural diagram of an audio encoder 300 provided in another embodiment of the present invention.

The time-domain audio signal can be input into the audio encoder 300 provided in the embodiment of the present invention in units of frames. After encoding by the audio encoder 300, the input audio frame can be compressed into a relatively small bitstream. This bitstream can be used for storage or transmission purposes, and can be restored to the original time-domain audio frame through an audio decoder.

Among them, the audio encoder 300 in this embodiment may include multiple sub-encoders, specifically, at least one sub-encoder is a sub-encoder based on linear prediction (for convenience, the sub-encoder based on linear prediction may be referred to as a Class A sub-encoder hereinafter), and at least one sub-encoder is a sub-encoder not based on linear prediction (for convenience, the sub-encoder not based on linear prediction may be referred to as a Class B encoder hereinafter).

As shown in FIG3 - a , the audio encoder 300 includes a selector 301 , a class A sub-encoder 302 , a class B sub-encoder 303 and a controlled router 304 .

The selector 301 is configured to estimate a reference linear prediction efficiency for the current audio frame, determine an audio encoder that matches the reference linear prediction efficiency for the current audio frame, and send a routing control signal to the controlled router 304 to control the controlled router 304 to output the current audio frame input to the controlled router 304 to the audio encoder (e.g., the Class A sub-encoder 302 or the Class B sub-encoder 303) that matches the reference linear prediction efficiency for the current audio frame. The Class A sub-encoder 302 or the Class B sub-encoder 303 is configured to perform audio encoding on the input current audio frame and output an encoded audio signal. For example, the Class A sub-encoder 302 may be a TCX encoder, and the Class B sub-encoder 302 may be a GAC encoder. For example, the Class B sub-encoder 302 may be an MDCT encoder.

In some embodiments of the present invention, as shown in FIG3 - b , a classifier 305 and a sub-encoder 306 may be further added based on the audio encoder 300 of the architecture shown in FIG3 - a .

Classifier 305 is configured to determine whether the current audio frame is a speech audio frame. If so, it sends a routing control signal to controlled router 304 to control controlled router 304 to output the current audio frame input to sub-encoder 306. Encoder 306 is a sub-encoder suitable for encoding speech audio frames, such as an ACELP encoder. Encoder 306 is configured to perform audio encoding on the input current audio frame and output an encoded audio signal.

In some embodiments of the present invention, as shown in FIG3-c , the selector 301 may include a decision unit 3013, a first estimation unit 3011, and a second estimation unit 3022. The reference linear prediction efficiency of the audio frame includes a reference long-time linear prediction efficiency and a reference short-time linear prediction efficiency of the audio frame.

The first estimation unit 3011 is used to estimate the reference long-term linear prediction efficiency of the current audio frame.

The second estimation unit 3012 is configured to estimate a reference short-time linear prediction efficiency of the current audio frame.

The decision unit 3013 is configured to determine that the audio coding method that matches the reference linear prediction efficiency of the current audio frame is a non-linear prediction-based audio coding method if the reference long-time linear prediction efficiency of the current audio frame estimated by the first estimation unit 3011 is less than a first threshold value, and/or the reference short-time linear prediction efficiency of the current audio frame estimated by the second estimation unit 3012 is less than a second threshold value, and to send a routing control signal to the controlled router 304 to control the controlled router 304 to output the current audio frame input into the controlled router 304 to the sub-class B sub-coding method. 303; if the reference long-time linear prediction efficiency of the current audio frame estimated by the first estimation unit 3011 is greater than or equal to the first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame estimated by the second estimation unit 3012 is greater than or equal to the second threshold, then it is determined that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction, and a routing control signal is sent to the controlled router 304 to control the controlled router 304 to output the current audio frame input into the controlled router 304 to the sub-B class subencoder 302.

In some embodiments of the present invention, as shown in FIG. 3 - d and FIG. 3 - e , the selector 301 also does not include the first estimation unit 3011 or the second estimation unit 3012 .

In the architecture shown in Figure 3-d, the judgment unit 3013 can be used to determine that the audio coding method matching the reference linear prediction efficiency of the above-mentioned current audio frame estimated by the first estimation unit 3011 is an audio coding method not based on linear prediction if the reference long-term linear prediction efficiency of the above-mentioned current audio frame estimated by the first estimation unit 3011 is less than the first threshold, and send a routing control signal to the controlled router 304 to control the controlled router 304 to output the current audio frame input into the controlled router 304 to the sub-B class sub-encoder 303; if the reference long-term linear prediction efficiency of the above-mentioned current audio frame estimated by the first estimation unit 3011 is greater than or equal to the first threshold, then determine that the audio coding method matching the reference linear prediction efficiency of the above-mentioned current audio frame is an audio coding method based on linear prediction, and send a routing control signal to the controlled router 304 to control the controlled router 304 to output the current audio frame input into the controlled router 304 to the sub-B class sub-encoder 302.

In the architecture shown in Figure 3-e, the judgment unit 3013 can be used to determine that the audio coding method matching the reference linear prediction efficiency of the above-mentioned current audio frame estimated by the second estimation unit 3012 is an audio coding method not based on linear prediction if the reference short-time linear prediction efficiency of the above-mentioned current audio frame estimated by the second estimation unit 3012 is less than the second threshold, and send a routing control signal to the controlled router 304 to control the controlled router 304 to output the current audio frame input into the controlled router 304 to the sub-B class sub-encoder 303; if the reference short-time linear prediction efficiency of the above-mentioned current audio frame estimated by the second estimation unit 3012 is greater than or equal to the second threshold, then determine that the audio coding method matching the reference linear prediction efficiency of the above-mentioned current audio frame is an audio coding method based on linear prediction, and send a routing control signal to the controlled router 304 to control the controlled router 304 to output the current audio frame input into the controlled router 304 to the sub-B class sub-encoder 302.

In some embodiments of the present invention, as shown in FIG3-f, based on the audio encoder 300 of the architecture shown in FIG3-c, the audio encoder 300 may further include a pre-processor 3014 for obtaining a linear prediction residual of the current audio frame. The pre-processor 3014 may be specifically used to filter the time domain signal of the current audio frame using an analysis filter A(Z) to obtain a linear prediction residual R of the current audio frame, wherein the filter coefficient of the filter A(Z) is the linear prediction coefficient of the current audio frame.

The first estimation unit 3011 is specifically configured to obtain, based on the linear prediction residual of the current audio frame and the first historical linear prediction signal, a correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal, and obtain, based on a mapping relationship between the correlation and the long-term linear prediction efficiency of the current audio frame, a long-term linear prediction efficiency of the current audio frame having a mapping relationship with the calculated correlation, wherein the first historical linear prediction signal is a first historical linear prediction excitation or a first historical linear prediction residual; the first historical linear prediction residual is a linear prediction residual of a historical audio frame of the current audio frame (for example, the first historical linear prediction residual may be a linear prediction residual of a historical audio frame of the current audio frame having the same or similar duration as the current audio frame and being the current audio frame). The first historical linear prediction excitation may be a linear prediction residual of a historical audio frame of a frequency frame, or the first historical linear prediction residual may be a linear prediction residual of a portion of a continuous audio signal of two adjacent historical audio frames having the same or similar duration as the current audio frame, and being the current audio frame), and the first historical linear prediction excitation may be a linear prediction excitation of a historical audio frame of the current audio frame (for example, the first historical linear prediction excitation may be a linear prediction excitation of a historical audio frame having the same or similar duration as the current audio frame, and being the current audio frame, or the first historical linear prediction excitation may be a linear prediction excitation of a portion of a continuous audio signal of two adjacent historical audio frames having the same or similar duration as the current audio frame).

In some embodiments of the present invention, as shown in FIG3-g , based on the audio encoder 300 of the architecture shown in FIG3-f , the audio encoder 300 may further include a buffer 308 . The buffer 308 may cache the linear prediction excitation or linear prediction residual of each audio frame or a portion of an audio frame, so as to serve as a historical linear prediction excitation or historical linear prediction residual that may be used in a possible next audio frame, thereby calculating its correlation with the linear prediction residual of the next audio frame. The first estimation unit 3011 may obtain the first historical linear prediction signal from the buffer 308 .

In some embodiments of the present invention, as shown in FIG3-h , the historical linear prediction excitation or historical linear prediction residual cached in the buffer 308 may come from the local audio decoder 311. The local audio decoder 311 may decode and output the encoded audio frames encoded by the Class A sub-encoder 302 and the Class B sub-encoder 303, and the linear predictor 312 may perform linear prediction on the time-domain audio frames output by the local audio decoder 311 to obtain the linear prediction residual or linear prediction excitation of the audio frames.

In some embodiments of the present invention, as shown in Figure 3-i, the historical linear prediction excitation cached by the buffer 308 may also come from the Class A sub-encoder 302. The Class A sub-encoder 302 will obtain the linear prediction excitation of the audio frame during the process of encoding the audio frame. The Class A sub-encoder 302 can output the obtained linear prediction excitation of the audio frame to the buffer 308 for caching.

In some embodiments of the present invention, the first historical linear prediction excitation or the first historical linear prediction residual used by the first estimation unit 3011 to estimate the long-term linear prediction efficiency of the current audio frame may be determined based on the pitch of the current audio frame. For example, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction excitations cached in the buffer 308 and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction excitation cached in the buffer 308 and the linear prediction residual of the current audio frame. For example, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction residuals cached in the buffer 308 and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction residual cached in the buffer 308 and the linear prediction residual of the current audio frame.

The audio encoder 300 may be any device that needs to collect, store or transmit audio signals, such as a mobile phone, a tablet computer, a personal computer, a laptop computer, etc.

4 , which is a schematic structural diagram of an audio encoder 400 provided by another embodiment of the present invention, wherein the audio encoder 400 may include an estimating unit 410 , a determining unit 420 , and an encoding unit 430 .

The estimation unit 410 is configured to estimate a reference linear prediction efficiency of the current audio frame.

The determining unit 420 is configured to determine an audio coding mode that matches the reference linear prediction efficiency of the current audio frame estimated by the estimating unit 410 .

The encoding unit 430 is configured to perform audio encoding on the current audio frame according to the audio encoding mode that matches the reference linear prediction efficiency of the current audio frame determined by the determining unit 420 .

In various embodiments of the present invention, a reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. A linear prediction result of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) refers to a linear prediction value of the audio frame. A higher reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree to which the audio frame can be linearly predicted.

In some embodiments of the present invention, the above-mentioned reference linear prediction efficiency includes at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the above-mentioned reference comprehensive linear prediction efficiency is obtained based on the above-mentioned reference long-time linear prediction efficiency and the above-mentioned reference short-time linear prediction efficiency.

For example, the reference long-term linear prediction efficiency of the current audio frame may be obtained based on the long-term linear prediction efficiency of the current audio frame. The reference short-term linear prediction efficiency of the current audio frame may be obtained based on the short-term linear prediction efficiency of the current audio frame. The reference comprehensive linear prediction efficiency of the current audio frame may be obtained based on the long-term linear prediction efficiency and the short-term linear prediction efficiency of the current audio frame.

It can be understood that the value range of the reference linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x1 (x1 is a positive number). Among them, the value range of the reference long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x2 (x2 is a positive number). The value range of the reference short-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x3 (x3 is a positive number). Among them, the value range of the reference comprehensive linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x4 (x4 is a positive number). Among them, the value range of the long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x5 (x5 is a positive number). The short-term linear prediction efficiency may range from 0 to 1 (i.e., 0% to 100%), or may range from 0 to x6 (x6 is a positive number). Here, x1, x2, x3, x4, x5, or x6 may be, for example, 0.5, 0.8, 1.5, 2, 5, 10, 50, 100, or other positive numbers.

In some embodiments of the present invention, the estimation unit may be specifically configured to estimate the reference linear prediction efficiency of the current audio frame when the current audio frame is a non-speech audio frame.

In some embodiments of the present invention, a reference comprehensive linear prediction efficiency for an audio frame (e.g., a current audio frame or another audio frame) is obtained based on a reference long-term linear prediction efficiency and a reference short-term linear prediction efficiency for the audio frame. The reference comprehensive linear prediction efficiency for the current audio frame may be, for example, the sum, weighted sum (wherein the weight corresponding to the weighted sum may be set as needed, such as 0.5, 1.0, 2.0, 3.0, 5.0, 10.0, or other values), or an average value of the reference long-term linear prediction efficiency and the reference short-term linear prediction efficiency for the current audio frame. Of course, the reference comprehensive linear prediction efficiency for the current audio frame may also be obtained using other algorithms based on the reference long-term linear prediction efficiency and the reference short-term linear prediction efficiency for the current audio frame.

It is understandable that the reference linear prediction efficiency of the current audio frame may include different types of linear prediction efficiencies, and the specific manner in which the determination unit 420 determines the audio encoding method that matches the reference linear prediction efficiency of the current audio frame may also be different.

Some possible implementation methods are given below.

In some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, the determination unit 420 may be specifically configured to: if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold, determine that the audio coding method matching the reference linear prediction efficiency of the current audio frame is an audio coding method not based on linear prediction.

In some further embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, the determination unit 420 may be specifically configured to: if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a second threshold, determine that the audio coding method that matches the reference linear prediction efficiency of the current audio frame is an audio coding method based on linear prediction.

In some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes a reference long-time linear prediction efficiency of the current audio frame and a reference short-time linear prediction efficiency of the current audio frame, the determination unit 420 may be specifically configured to: if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold, determine that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode that is not based on linear prediction; if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to the first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to the second threshold, determine that the audio coding mode that matches the reference linear prediction efficiency of the current audio frame is an audio coding mode that is based on linear prediction.

In some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, the determination unit 420 may be specifically used to: if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a third threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

In some further embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, the determination unit 420 may be specifically configured to: if the reference long-time linear prediction efficiency of the current audio frame is less than a fourth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method that is not based on linear prediction.

In some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference long-time linear prediction efficiency of the above-mentioned current audio frame, the determination unit 420 can be specifically used to: if the reference long-time linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a third threshold, determine that the audio encoding method matching the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference long-time linear prediction efficiency of the above-mentioned current audio frame is less than a fourth threshold, determine that the audio encoding method matching the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

In some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, the determination unit 420 is specifically used to: determine a first linear prediction efficiency interval in which the reference long-time linear prediction efficiency of the current audio frame falls, and determine a first audio coding method having a mapping relationship with the first linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio coding method based on linear prediction, wherein the first audio coding method is an audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the first audio coding method is an audio coding method based on linear prediction or an audio coding method not based on linear prediction.

In some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, then the above-mentioned determination unit 420 is specifically used to: if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to the fifth threshold, then determine that the audio encoding method matching the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction.

In some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, the determination unit 420 is specifically used to: if the reference short-time linear prediction efficiency of the current audio frame is less than a fifth threshold, determine that the audio encoding method matching the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

In some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, the determination unit 420 is specifically used to: if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a fifth threshold, determine that the audio encoding method matching the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is less than the fifth threshold, determine that the audio encoding method matching the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

In some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, the determination unit 420 is specifically used to: determine a second linear prediction efficiency interval into which the reference short-time linear prediction efficiency of the current audio frame falls, and determine a second audio coding method having a mapping relationship with the second linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio coding method based on linear prediction, wherein the second audio coding method is an audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the second audio coding method is an audio coding method based on linear prediction or an audio coding method not based on linear prediction.

In other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, the determination unit 420 is specifically used to: if the reference comprehensive linear prediction efficiency of the current audio frame is greater than or equal to a sixth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

In other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, the determination unit 420 is specifically used to: if the reference comprehensive linear prediction efficiency of the current audio frame is less than a sixth threshold, determine that the audio encoding method matching the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

In some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame, then the above-mentioned determination unit 420 is specifically used to: if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a sixth threshold, then determine that the audio encoding method matching the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is less than the sixth threshold, then determine that the audio encoding method matching the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

In other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, the determination unit 420 is specifically used to: determine the third linear prediction efficiency interval into which the reference comprehensive linear prediction efficiency of the current audio frame falls, and determine a third audio coding method having a mapping relationship with the third linear prediction efficiency interval based on a mapping relationship between the linear prediction efficiency interval and the audio coding method based on linear prediction, the third audio coding method being an audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the third audio coding method being an audio coding method based on linear prediction or an audio coding method not based on linear prediction.

In some embodiments of the present invention, audio coding methods based on linear prediction may include ACELP coding, TCX, etc. Audio coding methods not based on linear prediction may include GAC, which may include MDCT coding or DCT coding, for example.

It can be understood that the specific values of the various thresholds mentioned in the above examples (such as the first threshold, the second threshold, the third threshold, the fourth threshold, the fifth threshold, and the sixth threshold) can be set as needed or according to the application environment and scenario. For example, if the reference long-term linear prediction efficiency of the current audio frame ranges from 0 to 1, the first threshold can be 0.2, 0.5, 0.6, 0.8, etc., and if the reference short-term linear prediction efficiency of the current audio frame ranges from 0 to 1, the second threshold can be 0.3, 0.3, 0.6 or 0.8, etc. The same applies to other scenarios. Furthermore, the values of the various thresholds can be dynamically and adaptively adjusted as needed.

It is understandable that the estimation unit 410 may estimate the different types of linear prediction efficiencies included in the reference linear prediction efficiency of the current audio frame in different ways.

In some embodiments of the present invention, in terms of estimating the reference long-time linear prediction efficiency of the current audio frame, the estimation unit 410 is specifically used to: estimate the long-time linear prediction efficiency of the current audio frame, the long-time linear prediction efficiency of the current audio frame being the reference long-time linear prediction efficiency of the current audio frame.

In some other embodiments of the present invention, in terms of estimating the reference long-time linear prediction efficiency of the above-mentioned current audio frame, the estimation unit 410 is specifically used to: estimate the long-time linear prediction efficiency of the current audio frame; obtain the linear prediction efficiencies of N1 historical audio frames of the above-mentioned current audio frame; calculate the linear prediction efficiencies of the above-mentioned N1 historical audio frames and the first statistical value of the long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N1 is a positive integer, and the above-mentioned first statistical value is the reference long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the linear prediction efficiency of each historical audio frame in the N11 historical audio frames is at least one of the following linear prediction efficiency: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; the comprehensive linear prediction efficiency of the above-mentioned each historical audio frame is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of the above-mentioned each historical audio frame, and the above-mentioned N11 historical audio frames are a subset of the above-mentioned N1 historical audio frames. Among them, the first statistical value of the calculated linear prediction efficiency of the above N1 historical audio frames and the long-time linear prediction efficiency of the above current audio frame can be, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the above N1 historical audio frames and the long-time linear prediction efficiency of the above current audio frame.

In some other embodiments of the present invention, in terms of estimating the reference long-time linear prediction efficiency of the above-mentioned current audio frame, the estimation unit 410 is specifically used to: estimate the long-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiencies of N2 historical audio frames of the above-mentioned current audio frame; calculate the reference linear prediction efficiencies of the above-mentioned N2 historical audio frames and the second statistical value of the long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N2 is a positive integer, wherein the above-mentioned second statistical value is the reference long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the reference linear prediction efficiency of each historical audio frame in the N21 historical audio frames is at least one of the following linear prediction efficiencies: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of each historical audio frame is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of each historical audio frame, and the above-mentioned N21 historical audio frames are a subset of the above-mentioned N2 historical audio frames. Among them, the second statistical value of the reference linear prediction efficiency of the above N2 historical audio frames and the long-time linear prediction efficiency of the above current audio frame is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the reference linear prediction efficiency of the above N2 historical audio frames and the long-time linear prediction efficiency of the above current audio frame.

In some other embodiments of the present invention, in terms of estimating the reference long-time linear prediction efficiency of the above-mentioned current audio frame, the estimation unit 410 is specifically used to: estimate the long-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of N4 historical audio frames of the above-mentioned current audio frame, and obtain the linear prediction efficiency of N3 historical audio frames of the above-mentioned current audio frame; calculate the third statistical value of the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N3 and the above-mentioned N4 are positive integers, the above-mentioned third statistical value is the reference long-time linear prediction efficiency of the above-mentioned current audio frame, and the linear prediction efficiency of each historical audio frame in the N31 historical audio frames is at least one of the following linear prediction efficiencies : long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; wherein, the reference linear prediction efficiency of each historical audio frame in the N41 historical audio frames is at least one of the following linear prediction efficiency: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the above-mentioned N31 historical audio frames are a subset of the above-mentioned N3 historical audio frames, the above-mentioned N41 historical audio frames are a subset of the above-mentioned N4 historical audio frames, the comprehensive linear prediction efficiency of each of the above-mentioned historical audio frames is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each of the above-mentioned historical audio frames, and the reference comprehensive linear prediction efficiency of each of the above-mentioned historical audio frames is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each of the above-mentioned historical audio frames. Among them, the third statistical value calculated of the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted average of the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame.

In some embodiments of the present invention, in terms of estimating the reference short-time linear prediction efficiency of the above-mentioned current audio frame, the above-mentioned estimation unit 410 can be specifically used to: estimate the short-time linear prediction efficiency of the current audio frame, wherein the short-time linear prediction efficiency of the above-mentioned current audio frame is the reference short-time linear prediction efficiency of the above-mentioned current audio frame.

In some other embodiments of the present invention, in terms of estimating the reference short-time linear prediction efficiency of the above-mentioned current audio frame, the above-mentioned estimation unit 410 can be specifically used to: estimate the short-time linear prediction efficiency of the current audio frame; obtain the linear prediction efficiencies of N5 historical audio frames of the above-mentioned current audio frame; calculate the fourth statistical value of the linear prediction efficiencies of the above-mentioned N5 historical audio frames and the short-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N5 is a positive integer, and the above-mentioned fourth statistical value is the reference short-time linear prediction efficiency of the above-mentioned current audio frame, wherein the linear prediction efficiency of each historical audio frame in the N51 historical audio frames is at least one of the following linear prediction efficiency: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency, and the comprehensive linear prediction efficiency of each of the above-mentioned historical audio frames is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each of the above-mentioned historical audio frames, and the above-mentioned N51 historical audio frames are a subset of the above-mentioned N5 historical audio frames. The fourth statistical value of the linear prediction efficiency of the above N5 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the linear prediction efficiency of the above N5 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

In some other embodiments of the present invention, in terms of estimating the reference short-time linear prediction efficiency of the above-mentioned current audio frame, the above-mentioned estimation unit 410 can be specifically used to: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiencies of N6 historical audio frames of the above-mentioned current audio frame; calculate the fifth statistical value of the reference linear prediction efficiencies of the above-mentioned N6 historical audio frames and the short-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N6 is a positive integer, and the above-mentioned fifth statistical value is the reference short-time linear prediction efficiency of the above-mentioned current audio frame, wherein the reference linear prediction efficiency of each historical audio frame in the N61 historical audio frames is at least one of the following linear prediction efficiency: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of each historical audio frame is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of each historical audio frame, and the above-mentioned N61 historical audio frames are a subset of the above-mentioned N6 historical audio frames. Among them, the fifth statistical value of the reference linear prediction efficiency of the above N6 historical audio frames and the short-time linear prediction efficiency of the above current audio frame calculated by the estimation unit 410 may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the reference linear prediction efficiency of the above N6 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

In some other embodiments of the present invention, in terms of estimating the reference short-time linear prediction efficiency of the above-mentioned current audio frame, the above-mentioned estimation unit 410 can be specifically used to: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of N8 historical audio frames of the above-mentioned current audio frame; obtain the linear prediction efficiency of N7 historical audio frames of the above-mentioned current audio frame; calculate the sixth statistical value of the linear prediction efficiency of the above-mentioned N7 historical audio frames, the reference linear prediction efficiency of the above-mentioned N8 historical audio frames and the short-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N7 and the above-mentioned N8 are positive integers, the above-mentioned sixth statistical value is the reference short-time linear prediction efficiency of the above-mentioned current audio frame, and the linear prediction efficiency of each historical audio frame in the N71 historical audio frames is at least one of the following linear prediction efficiencies : long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency, the reference linear prediction efficiency of each historical audio frame in the N81 historical audio frames is at least one of the following linear prediction efficiencies: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the comprehensive linear prediction efficiency of each historical audio frame is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each historical audio frame, wherein the reference comprehensive linear prediction efficiency of each historical audio frame is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each historical audio frame, wherein the N71 historical audio frames are a subset of the N7 historical audio frames, and the N81 historical audio frames are a subset of the N8 historical audio frames. Among them, the sixth statistical value of the calculated linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

In some embodiments of the present invention, the linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. The linear prediction result of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) refers to the linear prediction value of the audio frame. A higher linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree of linear prediction for the audio frame.

In some embodiments of the present invention, in the aspect of estimating the short-time linear prediction efficiency of the current audio frame, the estimating unit 410 is specifically configured to obtain the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame.

In some embodiments of the present invention, regarding the aspect of obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame, the estimation unit 410 may be specifically configured to calculate an energy change rate before and after the short-time linear prediction of the current audio frame is performed, wherein the energy change rate is the short-time linear prediction efficiency of the current audio frame, or the short-time linear prediction efficiency of the current audio frame is obtained by transforming the energy change rate, wherein the energy of the current audio frame after the short-time linear prediction is the energy of the linear prediction residual of the current audio frame. For example, a mapping relationship may be established between the energy change rate and the short-time linear prediction efficiency of the current audio frame. Based on the mapping relationship between the energy change rate and the short-time linear prediction efficiency of the current audio frame, the short-time linear prediction efficiency of the current audio frame that has a mapping relationship with the calculated energy change rate may be obtained. Generally speaking, a greater energy change rate before and after the short-time linear prediction of the current audio frame is associated with the short-time linear prediction, indicating a higher short-time linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the energy change rate of the current audio frame before and after short-time linear prediction is calculated as a ratio of the energy of the current audio frame before short-time linear prediction to the energy of the linear prediction residual of the current audio frame. Generally speaking, a larger ratio obtained by dividing the energy of the current audio frame before short-time linear prediction by the energy of the linear prediction residual of the current audio frame indicates a higher short-time linear prediction efficiency for the current audio frame.

In some embodiments of the present invention, in terms of estimating the long-term linear prediction efficiency of the current audio frame, the estimation unit 410 may be specifically configured to: obtain the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the calculated linear prediction residual of the current audio frame, wherein the correlation is the long-term linear prediction efficiency of the current audio frame, or the long-term linear prediction efficiency of the current audio frame is obtained based on the correlation, wherein the first historical linear prediction signal is a first historical linear prediction excitation or a first historical linear prediction residual, and the first historical linear prediction residual is a linear prediction residual of a historical audio frame of the current audio frame (for example, the first historical linear prediction residual may be a linear prediction residual having a duration corresponding to the duration of the current audio frame). The first historical linear prediction excitation is the linear prediction excitation of the historical audio frame of the current audio frame (for example, the first historical linear prediction excitation can be the linear prediction excitation of the historical audio frame of the current audio frame, which is the same or similar in length as the current audio frame and is the linear prediction excitation of a historical audio frame of the current audio frame, or the first historical linear prediction excitation can be the linear prediction excitation of the historical audio frame of the current audio frame, which is the same or similar in length as the current audio frame and is the linear prediction excitation of a historical audio frame of the current audio frame).

There are various ways in which the estimation unit 410 obtains the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal.

In some embodiments of the present invention, in the aspect of obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on calculating the linear prediction residual of the current audio frame and the first historical linear prediction signal, the estimation unit 410 may be specifically configured to: calculate the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal;

Alternatively, multiplying the linear prediction residual of the current audio frame by a gain factor to obtain a gain linear prediction residual of the current audio frame, and calculating a correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal, wherein the calculated correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal;

or,

The first historical linear prediction signal is multiplied by a gain factor to obtain a first historical linear prediction signal after gain, and a correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal after gain is calculated, wherein the calculated correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal after gain is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

Generally speaking, a greater correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal indicates a higher long-term linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the first historical linear prediction excitation or the first historical linear prediction residual may be determined based on the pitch of the current audio frame. For example, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction excitations and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction excitation and the linear prediction residual of the current audio frame. For example, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction residuals and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction residual and the linear prediction residual of the current audio frame.

In some embodiments of the present invention, the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal can be, for example, a cross-correlation function value in the time domain and/or a cross-correlation function value in the frequency domain. Alternatively, the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal can be distortion in the time domain and/or distortion in the frequency domain. In some embodiments of the present invention, the distortion in the frequency domain can be the sum or weighted sum of distortions at K1 frequency points in the frequency domain, or the distortion in the frequency domain can be the sum or weighted sum of distortions at K2 subbands in the frequency domain, where K1 and K2 are positive integers. In some embodiments of the present invention, the weighting coefficient corresponding to the weighted sum of the distortions is a perceptual weighting coefficient reflecting a psychoacoustic model. Of course, the weighting coefficient corresponding to the weighted sum of the distortions can also be other weighting coefficients set based on actual needs. Testing has found that using a perceptual weighting coefficient helps make the calculated distortion more consistent with subjective quality, thereby improving performance.

Generally speaking, the larger the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the time domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the larger the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the frequency domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the smaller the distortion between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the frequency domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the smaller the distortion between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the time domain, the higher the long-term linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the first historical linear prediction excitation is a linear prediction excitation generated by audio encoding a historical audio frame of the current audio frame using a linear prediction-based encoding method.

In some embodiments of the present invention, the first historical linear prediction residual is obtained based on a time domain signal of a first historical audio frame of the current audio frame and linear prediction coefficients of the first historical audio frame, wherein the linear prediction coding coefficients of the first historical audio frame are quantized linear prediction coefficients or unquantized linear prediction coefficients. Since the quantized linear prediction coefficients generally contribute to the final quality in actual encoding and decoding processes, using the quantized linear prediction coefficients to calculate the linear prediction residual facilitates more accurate calculated correlation.

In some embodiments of the present invention, the linear prediction residual of the current audio frame is obtained based on the time domain signal of the current audio frame and the linear prediction coefficient of the current audio frame, wherein the linear prediction coefficient of the current audio frame is a quantized linear prediction coefficient or an unquantized linear prediction coefficient. Since the quantized linear prediction coefficients generally affect the final quality in actual encoding and decoding processes, using the quantized linear prediction coefficients to calculate the linear prediction residual facilitates more accurate calculated correlation.

In some embodiments of the present invention, the first historical linear prediction excitation is a superposition of an adaptive codebook excitation and a fixed codebook excitation, or the first historical linear prediction excitation is an adaptive codebook excitation.

It is understood that the functions of the various functional modules of the audio encoder 400 of this embodiment can be specifically implemented according to the methods in the above-mentioned method embodiments. The specific implementation process can be referred to the relevant description of the above-mentioned method embodiments and will not be repeated here. The audio encoder 400 can be any device that needs to collect, store, or transmit audio signals, such as a mobile phone, tablet computer, personal computer, laptop computer, etc.

Among them, for examples of values of various thresholds (such as the first threshold, the second threshold, etc.) and other parameters (such as N1, N11, N21, N2, etc.) involved in the embodiment of this device, please refer to the relevant value examples in the above method embodiment, and no further details will be given here.

It can be seen that in the technical solution of this embodiment, the audio encoder 400 first estimates the reference linear prediction efficiency of the current audio frame; determines the audio coding method that matches it through the estimated reference linear prediction efficiency of the above-mentioned current audio frame, and performs audio encoding on the above-mentioned current audio frame according to the determined audio coding method that matches it. Since the above-mentioned scheme does not need to perform the operation of fully encoding the current audio frame using each audio coding method required by the existing closed-loop selection mode in the process of determining the audio coding method, but instead determines the audio coding method to be selected through the reference linear prediction efficiency of the current audio frame, and the computational complexity of estimating the reference linear prediction efficiency of the current audio frame is usually much smaller than the computational complexity of fully encoding the current audio frame using each audio coding method, compared with the existing mechanism, the above-mentioned scheme of the embodiment of the present invention is conducive to reducing the computational complexity of audio coding, thereby reducing the overhead of audio coding.

Referring to FIG. 5 , FIG. 5 describes the structure of an encoder for decoding speech and audio code streams provided by another embodiment of the present invention. The encoder includes: at least one bus 501, at least one processor 502 connected to the bus 501, and at least one memory 503 connected to the bus 501.

In which, the processor 502 calls the code stored in the memory 503 through the bus 501 to estimate the reference linear prediction efficiency of the current audio frame; determines the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame; and audio encodes the above-mentioned current audio frame according to the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame.

In various embodiments of the present invention, a reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. A linear prediction result of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) refers to a linear prediction value of the audio frame. A higher reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree to which the audio frame can be linearly predicted.

In some embodiments of the present invention, the above-mentioned reference linear prediction efficiency includes at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the above-mentioned reference comprehensive linear prediction efficiency is obtained based on the above-mentioned reference long-time linear prediction efficiency and the above-mentioned reference short-time linear prediction efficiency.

The reference long-term linear prediction efficiency of the current audio frame may be obtained based on the long-term linear prediction efficiency of the current audio frame. The reference short-term linear prediction efficiency of the current audio frame may be obtained based on the short-term linear prediction efficiency of the current audio frame. The reference comprehensive linear prediction efficiency of the current audio frame may be obtained based on, for example, the long-term linear prediction efficiency and the short-term linear prediction efficiency of the current audio frame.

It can be understood that the value range of the reference linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x1 (x1 is a positive number). Among them, the value range of the reference long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x2 (x2 is a positive number). The value range of the reference short-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x3 (x3 is a positive number). Among them, the value range of the reference comprehensive linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x4 (x4 is a positive number). Among them, the value range of the long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x5 (x5 is a positive number). The short-term linear prediction efficiency may range from 0 to 1 (i.e., 0% to 100%), or may range from 0 to x6 (x6 is a positive number). Here, x1, x2, x3, x4, x5, or x6 may be, for example, 0.5, 0.8, 1.5, 2, 5, 10, 50, 100, or other positive numbers.

In some embodiments of the present invention, a predetermined mapping relationship may be established between the audio coding mode and the reference linear prediction efficiency of the audio frame. For example, different audio coding modes may correspond to different reference linear prediction efficiencies, or different audio coding modes may correspond to different reference linear prediction efficiency ranges. For example, an audio coding mode that matches the estimated reference linear prediction efficiency of the current audio frame may be determined from at least two audio coding modes.

In some embodiments of the present invention, before estimating the reference linear prediction efficiency of the current audio frame, the processor 502 may further be configured to call code stored in the memory 503 via the bus 501 to first determine whether the current audio frame is a speech audio frame. For example, estimating the reference linear prediction efficiency of the current audio frame may include, when the current audio frame is a non-speech audio frame, estimating the reference linear prediction efficiency of the current audio frame. Furthermore, the estimation of the reference linear prediction efficiency of the current audio frame may be performed without distinguishing whether the current audio frame is a speech audio frame.

In some embodiments of the present invention, the reference comprehensive linear prediction efficiency of the current audio frame may be, for example, the sum, weighted sum (wherein the weight corresponding to the weighted sum may be set as needed, and one weight may be, for example, 0.5, 1.0, 2.0, 3.0, 5.0, 10.0, or other values), or the average value of the reference long-term linear prediction efficiency and the reference short-term linear prediction efficiency of the current audio frame. Of course, the reference comprehensive linear prediction efficiency of the current audio frame may also be obtained using other algorithms based on the reference long-term linear prediction efficiency and the reference short-term linear prediction efficiency of the current audio frame.

It is understood that the reference linear prediction efficiency of the current audio frame may include different types of linear prediction efficiencies, and the specific manner in which the processor 502 determines the audio coding mode that matches the reference linear prediction efficiency of the current audio frame may also vary. Some possible embodiments are described below.

For example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then in the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to be specifically used for, if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then in the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to be specifically used for, if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a second threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then in the aspect of determining the audio coding method that matches the reference linear prediction efficiency of the current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically determine that, if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold, then the audio coding method that matches the reference linear prediction efficiency of the current audio frame is determined to be an audio coding method not based on linear prediction; if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to the first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to the second threshold, then the audio coding method that matches the reference linear prediction efficiency of the current audio frame is determined to be an audio coding method based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, then in the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to be specifically used for, if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a third threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference long-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to be specifically used for, if the reference long-time linear prediction efficiency of the above-mentioned current audio frame is less than a fourth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference long-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically be used for: if the reference long-time linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a third threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference long-time linear prediction efficiency of the above-mentioned current audio frame is less than a fourth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes a reference long-term linear prediction efficiency of the current audio frame, then in determining the audio coding mode that matches the reference linear prediction efficiency of the current audio frame, the processor 502 invokes code stored in the memory 503 via the bus 501 to specifically determine a first linear prediction efficiency interval within which the reference long-term linear prediction efficiency of the current audio frame falls, and based on a mapping relationship between linear prediction efficiency intervals and audio coding modes based on linear prediction, determine a first audio coding mode that has a mapping relationship with the first linear prediction efficiency interval, wherein the first audio coding mode is an audio coding mode that matches the reference linear prediction efficiency of the current audio frame, and the first audio coding mode is an audio coding mode based on linear prediction or an audio coding mode not based on linear prediction. Different linear prediction efficiency intervals correspond to different audio coding modes. For example, assuming there are three linear prediction efficiency intervals, which can be 0-30%, 30%-70%, and 70%-100%, respectively, if the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-30% (i.e., the first linear prediction efficiency interval is the linear prediction efficiency interval of 0-30%), the audio coding mode corresponding to the linear prediction efficiency interval of 0-30% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 30%-70% (i.e., the first linear prediction efficiency interval is the linear prediction efficiency interval of 30%-70%), the audio coding mode corresponding to the linear prediction efficiency interval of 30%-70% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between linear prediction efficiency intervals and linear prediction-based audio coding modes can be set according to the needs of different application scenarios.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to be specifically used for, if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to the fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to be specifically used for, if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is less than the fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically be used for: if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to the fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is less than the fifth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then in the aspect of determining the audio coding method that matches the reference linear prediction efficiency of the current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically determine the second linear prediction efficiency interval into which the reference short-time linear prediction efficiency of the current audio frame falls, and according to the mapping relationship between the linear prediction efficiency interval and the audio coding method based on linear prediction, determines the second audio coding method that has a mapping relationship with the second linear prediction efficiency interval or the audio coding method that is not based on linear prediction, wherein the second audio coding method is the audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the second audio coding method is the audio coding method based on linear prediction. For example, assuming there are three linear prediction efficiency intervals, namely 0-40%, 40%-60%, and 60%-100%, if the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-40% (i.e., the second linear prediction efficiency interval is the linear prediction efficiency interval of 0-40%), then the audio coding mode corresponding to the linear prediction efficiency interval of 0-40% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 40%-60% (i.e., the second linear prediction efficiency interval is the linear prediction efficiency interval of 40%-60%), then the audio coding mode corresponding to the linear prediction efficiency interval of 40%-60% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between linear prediction efficiency intervals and linear prediction-based audio coding modes can be set according to the needs of different application scenarios.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to be specifically used for, if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to be specifically used for, if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is less than the sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically be used for: if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is less than the sixth threshold, then determining that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then in the aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the current audio frame, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically determine the third linear prediction efficiency interval into which the reference comprehensive linear prediction efficiency of the current audio frame falls, and according to the mapping relationship between the linear prediction efficiency interval and the audio encoding method based on linear prediction, determines a third audio encoding method that has a mapping relationship with the third linear prediction efficiency interval or an audio encoding method that is not based on linear prediction, wherein the third audio encoding method is an audio encoding method that matches the reference linear prediction efficiency of the current audio frame, and the third audio encoding method is an audio encoding method based on linear prediction. For example, assuming there are three linear prediction efficiency intervals, which can be 0-50%, 50%-80%, and 80%-100%, respectively. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-50% (i.e., the third linear prediction efficiency interval is the linear prediction efficiency interval of 0-50%), the audio coding mode corresponding to the linear prediction efficiency interval of 0-50% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 50%-80% (i.e., the third linear prediction efficiency interval is the linear prediction efficiency interval of 50%-80%), the audio coding mode corresponding to the linear prediction efficiency interval of 50%-80% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between linear prediction efficiency intervals and linear prediction-based audio coding modes can be set according to the needs of different application scenarios.

In some embodiments of the present invention, audio coding methods based on linear prediction may include Algebraic Code Excited Linear Prediction (ACELP) coding, Transform Excited Coding (TCX), etc. Audio coding methods not based on linear prediction may include General Audio Coding (GAC), which may include, for example, Modified Discrete Cosine Transform (MDCT) coding or Discrete Cosine Transform (DCT) coding.

It can be understood that the specific values of the various thresholds mentioned in the above examples (such as the first threshold, the second threshold, the third threshold, the fourth threshold, the fifth threshold, and the sixth threshold) can be set as needed or according to the application environment and scenario. For example, if the reference long-term linear prediction efficiency of the current audio frame ranges from 0 to 1, the first threshold can be 0.2, 0.5, 0.6, 0.8, etc., and if the reference short-term linear prediction efficiency of the current audio frame ranges from 0 to 1, the second threshold can be 0.3, 0.3, 0.6 or 0.8, etc. The same applies to other scenarios. Furthermore, the values of the various thresholds can be dynamically and adaptively adjusted as needed.

It is understandable that the specific estimation methods of different types of linear prediction efficiencies included in the reference linear prediction efficiency of the current audio frame may be different. The following describes some possible embodiments.

For example, in some embodiments of the present invention, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically estimate the reference long-time linear prediction efficiency of the current audio frame in the following manner: estimating the long-time linear prediction efficiency of the current audio frame, the long-time linear prediction efficiency of the current audio frame being the reference long-time linear prediction efficiency of the current audio frame.

Alternatively, the processor 502 calls the code stored in the memory 503 through the bus 501, specifically for estimating the reference long-time linear prediction efficiency of the current audio frame in the following manner: estimating the long-time linear prediction efficiency of the current audio frame; obtaining the linear prediction efficiencies of N1 historical audio frames of the current audio frame; calculating the linear prediction efficiencies of the N1 historical audio frames and the first statistical value of the long-time linear prediction efficiency of the current audio frame, wherein the N1 is a positive integer, and the first statistical value is the reference long-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N11 historical audio frames is at least one of the following linear prediction efficiencies of each of the historical audio frames: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; the comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each of the historical audio frames, and the N11 historical audio frames are a subset of the N1 historical audio frames. Among them, the first statistical value of the calculated linear prediction efficiency of the above N1 historical audio frames and the long-time linear prediction efficiency of the above current audio frame can be, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the above N1 historical audio frames and the long-time linear prediction efficiency of the above current audio frame.

Alternatively, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically estimate the reference long-time linear prediction efficiency of the current audio frame in the following manner: estimate the long-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiencies of N2 historical audio frames of the current audio frame; calculate the reference linear prediction efficiencies of the N2 historical audio frames and a second statistical value of the long-time linear prediction efficiency of the current audio frame, wherein the N2 is a positive integer, and the second statistical value is the reference long-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each of the N21 historical audio frames is at least one of the following linear prediction efficiencies of each of the historical audio frames: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of each of the historical audio frames, and the N21 historical audio frames are a subset of the N2 historical audio frames. The second statistical value of the reference linear prediction efficiency of the above N2 historical audio frames and the long-time linear prediction efficiency of the above current audio frame is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the reference linear prediction efficiency of the above N2 historical audio frames and the long-time linear prediction efficiency of the above current audio frame.

Alternatively, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically estimate the reference long-time linear prediction efficiency of the current audio frame in the following manner: estimate the long-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of the N4 historical audio frames of the above-mentioned current audio frame, and obtain the linear prediction efficiency of the N3 historical audio frames of the above-mentioned current audio frame; calculate the third statistical value of the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N3 and the above-mentioned N4 are positive integers, the above-mentioned third statistical value is the reference long-time linear prediction efficiency of the above-mentioned current audio frame, and the linear prediction efficiency of each historical audio frame in the N31 historical audio frames is the following linear prediction efficiency of each of the above-mentioned historical audio frames At least one of the following: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency; wherein, the reference linear prediction efficiency of each historical audio frame in the N41 historical audio frames is at least one of the following linear prediction efficiencies of each historical audio frame: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency, and reference comprehensive linear prediction efficiency, wherein the N31 historical audio frames are a subset of the N3 historical audio frames, the N41 historical audio frames are a subset of the N4 historical audio frames, the comprehensive linear prediction efficiency of each historical audio frame is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each historical audio frame, and the reference comprehensive linear prediction efficiency of each historical audio frame is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each historical audio frame. The intersection of the N3 historical audio frames and the N4 historical audio frames may be an empty set or not an empty set. The third statistical value calculated for the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame.

For example, in some embodiments of the present invention, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically estimate the reference short-time linear prediction efficiency of the current audio frame in the following manner: estimating the short-time linear prediction efficiency of the current audio frame, wherein the short-time linear prediction efficiency of the current audio frame is the reference short-time linear prediction efficiency of the current audio frame.

Alternatively, the processor 502 calls the code stored in the memory 503 through the bus 501, specifically for estimating the reference short-time linear prediction efficiency of the current audio frame in the following manner: estimating the short-time linear prediction efficiency of the current audio frame; obtaining the linear prediction efficiencies of N5 historical audio frames of the current audio frame; calculating a fourth statistical value of the linear prediction efficiencies of the N5 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein the N5 is a positive integer, and the fourth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N51 historical audio frames is at least one of the following linear prediction efficiencies of each of the historical audio frames: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency, and the comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the long-time linear prediction efficiency and the short-time linear prediction efficiency of each of the historical audio frames, and the N51 historical audio frames are a subset of the N5 historical audio frames. Among them, the fourth statistical value of the calculated linear prediction efficiency of the above N5 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the linear prediction efficiency of the above N5 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

Alternatively, the processor 502 calls the code stored in the memory 503 through the bus 501, specifically for estimating the reference short-time linear prediction efficiency of the current audio frame in the following manner: estimating the short-time linear prediction efficiency of the current audio frame; obtaining the reference linear prediction efficiencies of N6 historical audio frames of the current audio frame; calculating a fifth statistical value of the reference linear prediction efficiencies of the N6 historical audio frames and the short-time linear prediction efficiency of the current audio frame, where N6 is a positive integer and the fifth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each of the N61 historical audio frames is at least one of the following linear prediction efficiencies of each of the historical audio frames: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of each of the historical audio frames is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of each of the historical audio frames, and the N61 historical audio frames are a subset of the N6 historical audio frames. Among them, the fifth statistical value of the reference linear prediction efficiency of the above N6 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted average of the reference linear prediction efficiency of the above N6 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

Alternatively, the processor 502 calls the code stored in the memory 503 through the bus 501 to specifically estimate the reference short-time linear prediction efficiency of the current audio frame in the following manner: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of the N8 historical audio frames of the above-mentioned current audio frame; obtain the linear prediction efficiency of the N7 historical audio frames of the above-mentioned current audio frame; calculate the sixth statistical value of the linear prediction efficiency of the above-mentioned N7 historical audio frames, the reference linear prediction efficiency of the above-mentioned N8 historical audio frames and the short-time linear prediction efficiency of the above-mentioned current audio frame, the above-mentioned N7 and the above-mentioned N8 are positive integers, the above-mentioned sixth statistical value is the reference short-time linear prediction efficiency of the above-mentioned current audio frame, and the linear prediction efficiency of each historical audio frame in the N71 historical audio frames is the following linear prediction efficiency of each of the above-mentioned historical audio frames At least one of the following: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; the reference linear prediction efficiency of each historical audio frame in the N81 historical audio frames is at least one of the following linear prediction efficiencies of each historical audio frame: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency; the comprehensive linear prediction efficiency of each historical audio frame is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each historical audio frame; wherein the reference comprehensive linear prediction efficiency of each historical audio frame is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each historical audio frame; the N71 historical audio frames are a subset of the N7 historical audio frames; and the N81 historical audio frames are a subset of the N8 historical audio frames. The intersection of the N7 historical audio frames and the N8 historical audio frames may be an empty set or not an empty set. Among them, the sixth statistical value of the calculated linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

In some embodiments of the present invention, the linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. The linear prediction result of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) refers to the linear prediction value of the audio frame. A higher linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree of linear prediction for the audio frame.

In some embodiments of the present invention, regarding the aspect of obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame, the processor 502 may invoke code stored in the memory 503 via the bus 501 to specifically calculate an energy change rate before and after the short-time linear prediction of the current audio frame, wherein the calculated energy change rate is the short-time linear prediction efficiency of the current audio frame, or the short-time linear prediction efficiency of the current audio frame is obtained by transforming the calculated energy change rate, wherein the energy of the current audio frame after the short-time linear prediction is the energy of the linear prediction residual of the current audio frame. For example, a mapping relationship may exist between the energy change rate and the short-time linear prediction efficiency of the current audio frame, and based on the mapping relationship between the energy change rate and the short-time linear prediction efficiency of the current audio frame, the short-time linear prediction efficiency of the current audio frame that has a mapping relationship with the calculated energy change rate may be obtained. Generally speaking, a greater energy change rate before and after the short-time linear prediction of the current audio frame indicates a higher short-time linear prediction efficiency of the current audio frame.

For example, the energy change rate of the current audio frame before and after short-time linear prediction may be a ratio of the energy of the current audio frame before short-time linear prediction to the energy of the linear prediction residual of the current audio frame, or the inverse of the ratio. Generally speaking, a larger ratio of the energy of the current audio frame before short-time linear prediction divided by the energy of the linear prediction residual of the current audio frame indicates a higher short-time linear prediction efficiency for the current audio frame.

In some embodiments of the present invention, in terms of estimating the long-term linear prediction efficiency of the current audio frame, the processor 502 may call the code stored in the memory 503 via the bus 501 to specifically be used for: obtaining, based on the linear prediction residual of the current audio frame and the first historical linear prediction signal, a correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal, where the correlation is the long-term linear prediction efficiency of the current audio frame, or the long-term linear prediction efficiency of the current audio frame is obtained based on the above-mentioned transformation. In which, the above-mentioned first historical linear prediction signal is the first historical linear prediction excitation or the first historical linear prediction residual; the above-mentioned first historical linear prediction residual is the linear prediction residual of the historical audio frame of the above-mentioned current audio frame (for example, the above-mentioned first historical linear prediction residual can be the linear prediction residual of a historical audio frame having the same or similar duration as the above-mentioned current audio frame and being the current audio frame, or the above-mentioned first historical linear prediction residual can be the linear prediction residual of a part of the continuous audio signal of two adjacent historical audio frames of the above-mentioned current audio frame, and being the current audio frame), and the above-mentioned first historical linear prediction excitation is the linear prediction excitation of the historical audio frame of the above-mentioned current audio frame (for example, the above-mentioned first historical linear prediction excitation can be the linear prediction excitation of a historical audio frame having the same or similar duration as the above-mentioned current audio frame and being the current audio frame, or the above-mentioned first historical linear prediction excitation can be the linear prediction excitation of a part of the continuous audio signal of two adjacent historical audio frames of the current audio frame). For example, there is a mapping relationship between the correlation and the long-term linear prediction efficiency of the audio frame. Based on the mapping relationship between the correlation and the long-term linear prediction efficiency of the audio frame, the long-term linear prediction efficiency of the current audio frame having a mapping relationship with the calculated correlation can be obtained.

There are various ways to obtain the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal.

For example, in the aspect of obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal, the processor 502 can call the code stored in the memory 503 through the bus 501 to specifically calculate the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

Alternatively, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame may include: multiplying the linear prediction residual of the current audio frame by a gain factor to obtain the gain linear prediction residual of the current audio frame, and calculating the correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal, wherein the calculated correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

Alternatively, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal may include: multiplying the first historical linear prediction signal by a gain factor to obtain the gained first historical linear prediction signal, and calculating the correlation between the linear prediction residual of the current audio frame and the gained first historical linear prediction signal, wherein the calculated correlation between the linear prediction residual of the current audio frame and the gained first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

The first historical linear prediction excitation or the first historical linear prediction residual may be determined based on the pitch of the current audio frame. For example, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction excitations and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction excitation and the linear prediction residual of the current audio frame. For example, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction residuals and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction residual and the linear prediction residual of the current audio frame.

Generally speaking, a greater correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal indicates a higher long-term linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the above-mentioned correlation is, for example, a cross-correlation function value in the time domain and/or a cross-correlation function value in the frequency domain, or the above-mentioned correlation may be a distortion in the time domain and/or a distortion in the frequency domain (wherein the distortion in the frequency domain may also be referred to as spectral distortion).

In some embodiments of the present invention, the above-mentioned distortion in the frequency domain may be the sum or weighted sum of the distortions of K1 frequency points in the frequency domain, or the above-mentioned distortion in the frequency domain may be the sum or weighted sum of the distortions on K2 sub-bands in the frequency domain, and the above-mentioned K1 and the above-mentioned K2 are positive integers.

Generally speaking, the larger the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the time domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the larger the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the frequency domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the smaller the distortion between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the frequency domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the smaller the distortion between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the time domain, the higher the long-term linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the weighting coefficient corresponding to the weighted sum of the distortion values is a perceptual weighting coefficient reflecting a psychoacoustic model. Of course, the weighting coefficient corresponding to the weighted sum of the distortion values can also be other weighting coefficients set based on actual needs. Testing has found that using a perceptual weighting coefficient helps make the calculated distortion more consistent with subjective quality, thereby improving performance.

In some embodiments of the present invention, the first historical linear prediction excitation may be a linear prediction excitation generated by audio encoding a historical audio frame of the current audio frame using a linear prediction-based encoding method.

In some embodiments of the present invention, the first historical linear prediction residual may be obtained based on a time domain signal of a first historical audio frame of the current audio frame and linear prediction coefficients of the first historical audio frame, wherein the linear prediction coding coefficients of the first historical audio frame are quantized linear prediction coefficients or unquantized linear prediction coefficients. Since the quantized linear prediction coefficients generally contribute to the final quality in actual encoding and decoding processes, using the quantized linear prediction coefficients to calculate the linear prediction residual facilitates more accurate calculated correlation.

In some embodiments of the present invention, the linear prediction residual of the current audio frame may be obtained based on the time domain signal of the current audio frame and the linear prediction coefficient of the current audio frame, wherein the linear prediction coefficient of the current audio frame may be a quantized linear prediction coefficient or an unquantized linear prediction coefficient. Since the quantized linear prediction coefficients generally contribute to the final quality in actual encoding and decoding processes, using the quantized linear prediction coefficients to calculate the linear prediction residual facilitates more accurate calculated correlation.

In some embodiments of the present invention, the first historical linear prediction excitation may be a superposition of an adaptive codebook excitation and a fixed codebook excitation, or the first historical linear prediction excitation may be an adaptive codebook excitation. Alternatively, the first historical linear prediction excitation may be another type of codebook excitation.

It is understood that the functions of the various functional modules of the audio encoder 500 of this embodiment can be specifically implemented according to the methods in the above-mentioned method embodiments. The specific implementation process can be referred to the relevant description of the above-mentioned method embodiments and will not be repeated here. The audio encoder 500 can be any device that needs to collect, store, or transmit audio signals, such as a mobile phone, tablet computer, personal computer, laptop computer, etc.

Among them, for examples of values of various thresholds (such as the first threshold, the second threshold, etc.) and other parameters (such as N1, N11, N21, N2, etc.) involved in the embodiment of this device, please refer to the relevant value examples in the above method embodiment, and no further details will be given here.

It can be seen that in the technical solution of this embodiment, the audio encoder 500 first estimates the reference linear prediction efficiency of the current audio frame; determines the audio coding method that matches it through the estimated reference linear prediction efficiency of the above-mentioned current audio frame, and performs audio encoding on the above-mentioned current audio frame according to the determined audio coding method that matches it. Since the above-mentioned scheme does not need to perform the operation of fully encoding the current audio frame using each audio coding method required by the existing closed-loop selection mode in the process of determining the audio coding method, but instead determines the audio coding method to be selected through the reference linear prediction efficiency of the current audio frame, and the computational complexity of estimating the reference linear prediction efficiency of the current audio frame is usually much smaller than the computational complexity of fully encoding the current audio frame using each audio coding method, compared with the existing mechanism, the above-mentioned scheme of the embodiment of the present invention is conducive to reducing the computational complexity of audio coding, thereby reducing the overhead of audio coding.

Referring to FIG6 , FIG6 is a block diagram of an audio encoder 600 according to another embodiment of the present invention. The audio encoder 600 may include: at least one processor 601, at least one network interface 604 or other user interface 603, a memory 605, and at least one communication bus 602. The communication bus 602 is used to implement communication between these components. The audio encoder 600 may optionally include a user interface 603, including a display (e.g., a touch screen, LCD, CRT, holographic imaging, or projector), a pointing device (e.g., a mouse, trackball, touch pad, or touch screen), a camera, and/or a sound pickup device.

The memory 602 may include a read-only memory and a random access memory, and provides instructions and data to the processor 601. A portion of the memory 602 may also include a non-volatile random access memory (NVRAM).

In some embodiments, the memory 605 stores the following elements, executable modules, or data structures, or a subset or extended set thereof:

The operating system 6051 includes various system programs for implementing various basic services and processing hardware-based tasks.

The application module 6052 includes various application programs for implementing various application services.

The application module 6052 includes but is not limited to a set estimation unit 410, a determination unit 420, and an encoding unit 430.

In an embodiment of the present invention, by calling a program or instruction stored in the memory 605, the processor 601 is used to estimate a reference linear prediction efficiency of the current audio frame; determine an audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame; and perform audio encoding on the above-mentioned current audio frame according to the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame.

In various embodiments of the present invention, a reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. A linear prediction result of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) refers to a linear prediction value of the audio frame. A higher reference linear prediction efficiency of an audio frame (such as a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree to which the audio frame can be linearly predicted.

In some embodiments of the present invention, the above-mentioned reference linear prediction efficiency includes at least one of the following linear prediction efficiencies: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency, and a reference comprehensive linear prediction efficiency, wherein the above-mentioned reference comprehensive linear prediction efficiency is obtained based on the above-mentioned reference long-time linear prediction efficiency and the above-mentioned reference short-time linear prediction efficiency.

The reference long-term linear prediction efficiency of the current audio frame may be obtained based on the long-term linear prediction efficiency of the current audio frame. The reference short-term linear prediction efficiency of the current audio frame may be obtained based on the short-term linear prediction efficiency of the current audio frame. The reference comprehensive linear prediction efficiency of the current audio frame may be obtained based on, for example, the long-term linear prediction efficiency and the short-term linear prediction efficiency of the current audio frame.

It can be understood that the value range of the reference linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x1 (x1 is a positive number). Among them, the value range of the reference long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x2 (x2 is a positive number). The value range of the reference short-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x3 (x3 is a positive number). Among them, the value range of the reference comprehensive linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x4 (x4 is a positive number). Among them, the value range of the long-term linear prediction efficiency can be 0-1 (i.e., 0%-100%), or the value range can also be 0-x5 (x5 is a positive number). The short-term linear prediction efficiency may range from 0 to 1 (i.e., 0% to 100%), or may range from 0 to x6 (x6 is a positive number). Here, x1, x2, x3, x4, x5, or x6 may be, for example, 0.5, 0.8, 1.5, 2, 5, 10, 50, 100, or other positive numbers.

In some embodiments of the present invention, a predetermined mapping relationship may be established between the audio coding mode and the reference linear prediction efficiency of the audio frame. For example, different audio coding modes may correspond to different reference linear prediction efficiencies, or different audio coding modes may correspond to different reference linear prediction efficiency ranges. For example, an audio coding mode that matches the estimated reference linear prediction efficiency of the current audio frame may be determined from at least two audio coding modes.

In some embodiments of the present invention, before estimating the reference linear prediction efficiency of the current audio frame, the processor 601 may further be configured to determine whether the current audio frame is a speech audio frame by invoking a program or instruction stored in the memory 605. For example, estimating the reference linear prediction efficiency of the current audio frame may include estimating the reference linear prediction efficiency of the current audio frame when the current audio frame is a non-speech audio frame. Furthermore, estimating the reference linear prediction efficiency of the current audio frame may not distinguish whether the current audio frame is a speech audio frame before estimating the reference linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the reference comprehensive linear prediction efficiency of the current audio frame may be, for example, the sum, weighted sum (wherein the weight corresponding to the weighted sum may be set as needed, and one weight may be, for example, 0.5, 1.0, 2.0, 3.0, 5.0, 10.0, or other values), or the average value of the reference long-term linear prediction efficiency and the reference short-term linear prediction efficiency of the current audio frame. Of course, the reference comprehensive linear prediction efficiency of the current audio frame may also be obtained using other algorithms based on the reference long-term linear prediction efficiency and the reference short-term linear prediction efficiency of the current audio frame.

It is understood that the reference linear prediction efficiency of the current audio frame may include different types of linear prediction efficiencies, and the specific manner in which the processor 601 determines the audio coding mode that matches the reference linear prediction efficiency of the current audio frame may also vary. Some possible embodiments are described below.

For example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then in the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to, if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the current audio frame, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to, if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to a second threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame and the reference short-time linear prediction efficiency of the current audio frame, then in the aspect of determining the audio coding method that matches the reference linear prediction efficiency of the current audio frame, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to determine that the audio coding method that matches the reference linear prediction efficiency of the current audio frame is an audio coding method not based on linear prediction if the reference long-time linear prediction efficiency of the current audio frame is less than a first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is less than a second threshold; if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to the first threshold, and/or the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to the second threshold, then determine that the audio coding method that matches the reference linear prediction efficiency of the current audio frame is an audio coding method based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference long-time linear prediction efficiency of the current audio frame, then in the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to, if the reference long-time linear prediction efficiency of the current audio frame is greater than or equal to a third threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference long-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to, if the reference long-time linear prediction efficiency of the above-mentioned current audio frame is less than a fourth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference long-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to, if the reference long-time linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a third threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference long-time linear prediction efficiency of the above-mentioned current audio frame is less than a fourth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes a reference long-term linear prediction efficiency of the current audio frame, then in determining the audio coding mode that matches the reference linear prediction efficiency of the current audio frame, by invoking a program or instruction stored in the memory 605, the processor 601 may be specifically configured to determine a first linear prediction efficiency interval within which the reference long-term linear prediction efficiency of the current audio frame falls, and determine a first audio coding mode that has a mapping relationship with the first linear prediction efficiency interval based on a mapping relationship between linear prediction efficiency intervals and audio coding modes based on linear prediction, wherein the first audio coding mode is an audio coding mode that matches the reference linear prediction efficiency of the current audio frame, and the first audio coding mode is an audio coding mode based on linear prediction or an audio coding mode not based on linear prediction. Different linear prediction efficiency intervals correspond to different audio coding modes. For example, assuming there are three linear prediction efficiency intervals, which can be 0-30%, 30%-70%, and 70%-100%, respectively, if the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-30% (i.e., the first linear prediction efficiency interval is the linear prediction efficiency interval of 0-30%), the audio coding mode corresponding to the linear prediction efficiency interval of 0-30% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 30%-70% (i.e., the first linear prediction efficiency interval is the linear prediction efficiency interval of 30%-70%), the audio coding mode corresponding to the linear prediction efficiency interval of 30%-70% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between linear prediction efficiency intervals and linear prediction-based audio coding modes can be set according to the needs of different application scenarios.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to, if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to the fifth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to, if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is less than the fifth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference short-time linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to, if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a fifth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference short-time linear prediction efficiency of the above-mentioned current audio frame is less than the fifth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then in terms of determining the audio coding method that matches the reference linear prediction efficiency of the current audio frame, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to determine the second linear prediction efficiency interval in which the reference short-time linear prediction efficiency of the current audio frame falls, and according to the mapping relationship between the linear prediction efficiency interval and the audio coding method based on linear prediction, determine the second audio coding method that has a mapping relationship with the second linear prediction efficiency interval or the audio coding method that is not based on linear prediction, wherein the second audio coding method is the audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the second audio coding method is the audio coding method based on linear prediction. For example, assuming there are three linear prediction efficiency intervals, which can be 0-40%, 40%-60%, and 60%-100%, respectively. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-40% (i.e., the second linear prediction efficiency interval is the linear prediction efficiency interval of 0-40%), then the audio coding mode corresponding to the linear prediction efficiency interval of 0-40% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 40%-60% (i.e., the second linear prediction efficiency interval is the linear prediction efficiency interval of 40%-60%), then the audio coding mode corresponding to the linear prediction efficiency interval of 40%-60% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between linear prediction efficiency intervals and linear prediction-based audio coding modes can be set according to the needs of different application scenarios.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then in the above-mentioned determination of the audio encoding method that matches the reference linear prediction efficiency of the current audio frame, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to, if the reference comprehensive linear prediction efficiency of the current audio frame is greater than or equal to a sixth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the current audio frame is an audio encoding method based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to, if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is less than a sixth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method that is not based on linear prediction.

For another example, in some other embodiments of the present invention, if the reference linear prediction efficiency of the above-mentioned current audio frame includes the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame, then in the above-mentioned aspect of determining the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to, if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is greater than or equal to a sixth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method based on linear prediction; if the reference comprehensive linear prediction efficiency of the above-mentioned current audio frame is less than the sixth threshold, determine that the audio encoding method that matches the reference linear prediction efficiency of the above-mentioned current audio frame is an audio encoding method not based on linear prediction.

For another example, in some embodiments of the present invention, if the reference linear prediction efficiency of the current audio frame includes the reference comprehensive linear prediction efficiency of the current audio frame, then in terms of determining the audio encoding method that matches the reference linear prediction efficiency of the current audio frame, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to determine the third linear prediction efficiency interval into which the reference comprehensive linear prediction efficiency of the current audio frame falls, and according to the mapping relationship between the linear prediction efficiency interval and the audio encoding method based on linear prediction, determine a third audio encoding method that has a mapping relationship with the third linear prediction efficiency interval or an audio encoding method that is not based on linear prediction, wherein the third audio encoding method is an audio encoding method that matches the reference linear prediction efficiency of the current audio frame, and the third audio encoding method is an audio encoding method based on linear prediction. For example, assuming there are three linear prediction efficiency intervals, which can be 0-50%, 50%-80%, and 80%-100%, respectively. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 0-50% (i.e., the third linear prediction efficiency interval is the linear prediction efficiency interval of 0-50%), the audio coding mode corresponding to the linear prediction efficiency interval of 0-50% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. If the reference long-term linear prediction efficiency of the current audio frame falls within the linear prediction efficiency interval of 50%-80% (i.e., the third linear prediction efficiency interval is the linear prediction efficiency interval of 50%-80%), the audio coding mode corresponding to the linear prediction efficiency interval of 50%-80% can be determined to be the audio coding mode that matches the reference linear prediction efficiency of the current audio frame. The same applies to other scenarios. The mapping relationship between linear prediction efficiency intervals and linear prediction-based audio coding modes can be set according to the needs of different application scenarios.

In some embodiments of the present invention, audio coding methods based on linear prediction may include Algebraic Code Excited Linear Prediction (ACELP) coding, Transform Excited Coding (TCX), etc. Audio coding methods not based on linear prediction may include General Audio Coding (GAC), which may include, for example, Modified Discrete Cosine Transform (MDCT) coding or Discrete Cosine Transform (DCT) coding.

It can be understood that the specific values of the various thresholds mentioned in the above examples (such as the first threshold, the second threshold, the third threshold, the fourth threshold, the fifth threshold, and the sixth threshold) can be set as needed or according to the application environment and scenario. For example, if the reference long-term linear prediction efficiency of the current audio frame ranges from 0 to 1, the first threshold can be 0.2, 0.5, 0.6, 0.8, etc., and if the reference short-term linear prediction efficiency of the current audio frame ranges from 0 to 1, the second threshold can be 0.3, 0.3, 0.6 or 0.8, etc. The same applies to other scenarios. Furthermore, the values of the various thresholds can be dynamically and adaptively adjusted as needed.

It is understandable that the specific estimation methods of different types of linear prediction efficiencies included in the reference linear prediction efficiency of the current audio frame may be different. The following describes some possible embodiments.

For example, in some embodiments of the present invention, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to estimate the reference long-time linear prediction efficiency of the current audio frame in the following manner: estimating the long-time linear prediction efficiency of the current audio frame, the long-time linear prediction efficiency of the current audio frame being the reference long-time linear prediction efficiency of the current audio frame.

Alternatively, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to estimate the reference long-time linear prediction efficiency of the current audio frame in the following manner: estimate the long-time linear prediction efficiency of the current audio frame; obtain the linear prediction efficiencies of N1 historical audio frames of the above-mentioned current audio frame; calculate the linear prediction efficiencies of the above-mentioned N1 historical audio frames and the first statistical value of the long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N1 is a positive integer, and the above-mentioned first statistical value is the reference long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the linear prediction efficiency of each historical audio frame in the N11 historical audio frames is at least one of the following linear prediction efficiencies of each historical audio frame: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency; the comprehensive linear prediction efficiency of each historical audio frame is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each historical audio frame, and the above-mentioned N11 historical audio frames are a subset of the above-mentioned N1 historical audio frames. Among them, the first statistical value of the calculated linear prediction efficiency of the above N1 historical audio frames and the long-time linear prediction efficiency of the above current audio frame can be, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the above N1 historical audio frames and the long-time linear prediction efficiency of the above current audio frame.

Alternatively, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to estimate the reference long-time linear prediction efficiency of the current audio frame in the following manner: estimate the long-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiencies of N2 historical audio frames of the above-mentioned current audio frame; calculate the reference linear prediction efficiencies of the above-mentioned N2 historical audio frames and the second statistical value of the long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N2 is a positive integer, and the above-mentioned second statistical value is the reference long-time linear prediction efficiency of the above-mentioned current audio frame, wherein the reference linear prediction efficiency of each historical audio frame in the N21 historical audio frames is at least one of the following linear prediction efficiencies of the above-mentioned each historical audio frame: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of the above-mentioned each historical audio frame is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of the above-mentioned each historical audio frame, and the above-mentioned N21 historical audio frames are a subset of the above-mentioned N2 historical audio frames. The second statistical value of the reference linear prediction efficiency of the above N2 historical audio frames and the long-time linear prediction efficiency of the above current audio frame is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the reference linear prediction efficiency of the above N2 historical audio frames and the long-time linear prediction efficiency of the above current audio frame.

Alternatively, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to estimate the reference long-time linear prediction efficiency of the current audio frame in the following manner: estimate the long-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of the N4 historical audio frames of the above-mentioned current audio frame, and obtain the linear prediction efficiency of the N3 historical audio frames of the above-mentioned current audio frame; calculate the third statistical value of the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame, the above-mentioned N3 and the above-mentioned N4 are positive integers, the above-mentioned third statistical value is the reference long-time linear prediction efficiency of the above-mentioned current audio frame, and the linear prediction efficiency of each historical audio frame in the N31 historical audio frames is one of the following linear prediction efficiencies of the above-mentioned each historical audio frame At least one of: long-time linear prediction efficiency, short-time linear prediction efficiency, and comprehensive linear prediction efficiency; wherein, the reference linear prediction efficiency of each historical audio frame in the N41 historical audio frames is at least one of the following linear prediction efficiencies of each historical audio frame: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency, and reference comprehensive linear prediction efficiency, wherein the N31 historical audio frames are a subset of the N3 historical audio frames, the N41 historical audio frames are a subset of the N4 historical audio frames, the comprehensive linear prediction efficiency of each historical audio frame is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each historical audio frame, and the reference comprehensive linear prediction efficiency of each historical audio frame is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each historical audio frame. The intersection of the N3 historical audio frames and the N4 historical audio frames may be an empty set or not. The third statistical value calculated for the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame is, for example, the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted mean of the linear prediction efficiency of the above-mentioned N3 historical audio frames, the reference linear prediction efficiency of the above-mentioned N4 historical audio frames and the long-time linear prediction efficiency of the above-mentioned current audio frame.

For example, in some embodiments of the present invention, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to estimate the reference short-time linear prediction efficiency of the current audio frame in the following manner: estimating the short-time linear prediction efficiency of the current audio frame, wherein the short-time linear prediction efficiency of the current audio frame is the reference short-time linear prediction efficiency of the current audio frame.

Alternatively, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to estimate the reference short-time linear prediction efficiency of the current audio frame in the following manner: estimate the short-time linear prediction efficiency of the current audio frame; obtain the linear prediction efficiencies of N5 historical audio frames of the above-mentioned current audio frame; calculate the fourth statistical value of the linear prediction efficiencies of the above-mentioned N5 historical audio frames and the short-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N5 is a positive integer, and the above-mentioned fourth statistical value is the reference short-time linear prediction efficiency of the above-mentioned current audio frame, wherein the linear prediction efficiency of each historical audio frame in the N51 historical audio frames is at least one of the following linear prediction efficiencies of each historical audio frame: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency, and the comprehensive linear prediction efficiency of each historical audio frame is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each historical audio frame, and the above-mentioned N51 historical audio frames are a subset of the above-mentioned N5 historical audio frames. Among them, the fourth statistical value of the calculated linear prediction efficiency of the above N5 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the linear prediction efficiency of the above N5 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

Alternatively, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to estimate the reference short-time linear prediction efficiency of the current audio frame in the following manner: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiencies of N6 historical audio frames of the above-mentioned current audio frame; calculate the fifth statistical value of the reference linear prediction efficiencies of the above-mentioned N6 historical audio frames and the short-time linear prediction efficiency of the above-mentioned current audio frame, wherein the above-mentioned N6 is a positive integer, and the above-mentioned fifth statistical value is the reference short-time linear prediction efficiency of the above-mentioned current audio frame, wherein the reference linear prediction efficiency of each historical audio frame in the N61 historical audio frames is at least one of the following linear prediction efficiencies of each historical audio frame: a reference long-time linear prediction efficiency, a reference short-time linear prediction efficiency and a reference comprehensive linear prediction efficiency, wherein the reference comprehensive linear prediction efficiency of each historical audio frame is obtained based on the reference long-time linear prediction efficiency and the reference short-time linear prediction efficiency of each historical audio frame, and the above-mentioned N61 historical audio frames are a subset of the above-mentioned N6 historical audio frames. Among them, the fifth statistical value of the reference linear prediction efficiency of the above N6 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding mean or weighted average of the reference linear prediction efficiency of the above N6 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

Alternatively, by calling the program or instruction stored in the memory 605, the processor 601 can be specifically used to estimate the reference short-time linear prediction efficiency of the current audio frame in the following manner: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of the N8 historical audio frames of the above-mentioned current audio frame; obtain the linear prediction efficiency of the N7 historical audio frames of the above-mentioned current audio frame; calculate the sixth statistical value of the linear prediction efficiency of the above-mentioned N7 historical audio frames, the reference linear prediction efficiency of the above-mentioned N8 historical audio frames and the short-time linear prediction efficiency of the above-mentioned current audio frame, the above-mentioned N7 and the above-mentioned N8 are positive integers, the above-mentioned sixth statistical value is the reference short-time linear prediction efficiency of the above-mentioned current audio frame, and the linear prediction efficiency of each historical audio frame in the N71 historical audio frames is the following linear prediction efficiency of each of the above-mentioned historical audio frames At least one of: long-time linear prediction efficiency, short-time linear prediction efficiency and comprehensive linear prediction efficiency, the reference linear prediction efficiency of each historical audio frame in the N81 historical audio frames is at least one of the following linear prediction efficiencies of each historical audio frame: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency and reference comprehensive linear prediction efficiency, the comprehensive linear prediction efficiency of each historical audio frame is obtained based on the long-time linear prediction efficiency and short-time linear prediction efficiency of each historical audio frame, wherein the reference comprehensive linear prediction efficiency of each historical audio frame is obtained based on the reference long-time linear prediction efficiency and reference short-time linear prediction efficiency of each historical audio frame, the N71 historical audio frames are a subset of the N7 historical audio frames, and the N81 historical audio frames are a subset of the N8 historical audio frames. The intersection of the N7 historical audio frames and the N8 historical audio frames may be an empty set or not an empty set. Among them, the sixth statistical value of the calculated linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame may be the sum, weighted sum, geometric mean, arithmetic mean, sliding average or weighted average of the linear prediction efficiency of the above N7 historical audio frames, the reference linear prediction efficiency of the above N8 historical audio frames and the short-time linear prediction efficiency of the above current audio frame.

In some embodiments of the present invention, the linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) may be used to indicate the degree to which the audio frame can be linearly predicted. The linear prediction result of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) refers to the linear prediction value of the audio frame. A higher linear prediction efficiency (e.g., long-term linear prediction efficiency, short-term linear prediction efficiency) of an audio frame (e.g., a current audio frame or a historical audio frame of the current audio frame) indicates a higher degree of linear prediction for the audio frame.

In some embodiments of the present invention, in terms of obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame, the processor 601 may be specifically configured to, by invoking a program or instruction stored in the memory 605, calculate an energy change rate before and after performing short-time linear prediction on the current audio frame, wherein the calculated energy change rate is the short-time linear prediction efficiency of the current audio frame, or obtain the short-time linear prediction efficiency of the current audio frame by transforming the calculated energy change rate, wherein the energy of the current audio frame after performing short-time linear prediction is the energy of the linear prediction residual of the current audio frame. For example, a mapping relationship may be established between the energy change rate and the short-time linear prediction efficiency of the current audio frame. Based on the mapping relationship between the energy change rate and the short-time linear prediction efficiency of the current audio frame, the short-time linear prediction efficiency of the current audio frame that has a mapping relationship with the calculated energy change rate may be obtained. Generally speaking, a greater energy change rate before and after performing short-time linear prediction on the current audio frame indicates a higher short-time linear prediction efficiency of the current audio frame.

For example, the energy change rate of the current audio frame before and after short-time linear prediction may be a ratio of the energy of the current audio frame before short-time linear prediction to the energy of the linear prediction residual of the current audio frame, or the inverse of the ratio. Generally speaking, a larger ratio of the energy of the current audio frame before short-time linear prediction divided by the energy of the linear prediction residual of the current audio frame indicates a higher short-time linear prediction efficiency for the current audio frame.

In some embodiments of the present invention, in terms of estimating the long-term linear prediction efficiency of the current audio frame, the processor 601 may be specifically configured to, by calling a program or instruction stored in the memory 605, obtain, based on the linear prediction residual of the current audio frame and the first historical linear prediction signal, a correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal, where the correlation is the long-term linear prediction efficiency of the current audio frame, or the long-term linear prediction efficiency of the current audio frame is obtained based on the above-mentioned transformation. In which, the above-mentioned first historical linear prediction signal is the first historical linear prediction excitation or the first historical linear prediction residual; the above-mentioned first historical linear prediction residual is the linear prediction residual of the historical audio frame of the above-mentioned current audio frame (for example, the above-mentioned first historical linear prediction residual can be the linear prediction residual of a historical audio frame having the same or similar duration as the above-mentioned current audio frame and being the current audio frame, or the above-mentioned first historical linear prediction residual can be the linear prediction residual of a part of the continuous audio signal of two adjacent historical audio frames of the above-mentioned current audio frame, and being the current audio frame), and the above-mentioned first historical linear prediction excitation is the linear prediction excitation of the historical audio frame of the above-mentioned current audio frame (for example, the above-mentioned first historical linear prediction excitation can be the linear prediction excitation of a historical audio frame having the same or similar duration as the above-mentioned current audio frame and being the current audio frame, or the above-mentioned first historical linear prediction excitation can be the linear prediction excitation of a part of the continuous audio signal of two adjacent historical audio frames of the current audio frame). For example, there is a mapping relationship between the correlation and the long-term linear prediction efficiency of the audio frame. Based on the mapping relationship between the correlation and the long-term linear prediction efficiency of the audio frame, the long-term linear prediction efficiency of the current audio frame having a mapping relationship with the calculated correlation can be obtained.

There are various ways to obtain the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal.

For example, in the aspect of obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal, by calling the program or instructions stored in the memory 605, the processor 601 can be specifically used to calculate the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

Alternatively, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame may include: multiplying the linear prediction residual of the current audio frame by a gain factor to obtain the gain linear prediction residual of the current audio frame, and calculating the correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal, wherein the calculated correlation between the gain linear prediction residual of the current audio frame and the first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

Alternatively, obtaining the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal based on the linear prediction residual of the current audio frame and the first historical linear prediction signal may include: multiplying the first historical linear prediction signal by a gain factor to obtain the gained first historical linear prediction signal, and calculating the correlation between the linear prediction residual of the current audio frame and the gained first historical linear prediction signal, wherein the calculated correlation between the linear prediction residual of the current audio frame and the gained first historical linear prediction signal is the correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal.

The first historical linear prediction excitation or the first historical linear prediction residual may be determined based on the pitch of the current audio frame. For example, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction excitations and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction excitation and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction excitation and the linear prediction residual of the current audio frame. For example, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between other historical linear prediction residuals and the linear prediction residual of the current audio frame. Alternatively, the time-domain correlation between the first historical linear prediction residual and the linear prediction residual of the current audio frame is greater than or equal to the time-domain correlation between at least one other historical linear prediction residual and the linear prediction residual of the current audio frame.

Generally speaking, a greater correlation between the linear prediction residual of the current audio frame and the first historical linear prediction signal indicates a higher long-term linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the above-mentioned correlation is, for example, a cross-correlation function value in the time domain and/or a cross-correlation function value in the frequency domain, or the above-mentioned correlation may be a distortion in the time domain and/or a distortion in the frequency domain (wherein the distortion in the frequency domain may also be referred to as spectral distortion).

In some embodiments of the present invention, the above-mentioned distortion in the frequency domain may be the sum or weighted sum of the distortions of K1 frequency points in the frequency domain, or the above-mentioned distortion in the frequency domain may be the sum or weighted sum of the distortions on K2 sub-bands in the frequency domain, and the above-mentioned K1 and the above-mentioned K2 are positive integers.

Generally speaking, the larger the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the time domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the larger the cross-correlation function value between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the frequency domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the smaller the distortion between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the frequency domain, the higher the long-term linear prediction efficiency of the current audio frame. Generally speaking, the smaller the distortion between the linear prediction residual of the current audio frame and the first historical linear prediction signal in the time domain, the higher the long-term linear prediction efficiency of the current audio frame.

In some embodiments of the present invention, the weighting coefficient corresponding to the weighted sum of the distortion values is a perceptual weighting coefficient reflecting a psychoacoustic model. Of course, the weighting coefficient corresponding to the weighted sum of the distortion values can also be other weighting coefficients set based on actual needs. Testing has found that using a perceptual weighting coefficient helps make the calculated distortion more consistent with subjective quality, thereby improving performance.

In some embodiments of the present invention, the first historical linear prediction excitation may be a linear prediction excitation generated by audio encoding a historical audio frame of the current audio frame using a linear prediction-based encoding method.

In some embodiments of the present invention, the first historical linear prediction residual may be obtained based on a time domain signal of a first historical audio frame of the current audio frame and linear prediction coefficients of the first historical audio frame, wherein the linear prediction coding coefficients of the first historical audio frame are quantized linear prediction coefficients or unquantized linear prediction coefficients. Since the quantized linear prediction coefficients generally contribute to the final quality in actual encoding and decoding processes, using the quantized linear prediction coefficients to calculate the linear prediction residual facilitates more accurate calculated correlation.

In some embodiments of the present invention, the linear prediction residual of the current audio frame may be obtained based on the time domain signal of the current audio frame and the linear prediction coefficient of the current audio frame, wherein the linear prediction coefficient of the current audio frame may be a quantized linear prediction coefficient or an unquantized linear prediction coefficient. Since the quantized linear prediction coefficients generally contribute to the final quality in actual encoding and decoding processes, using the quantized linear prediction coefficients to calculate the linear prediction residual facilitates more accurate calculated correlation.

In some embodiments of the present invention, the first historical linear prediction excitation may be a superposition of an adaptive codebook excitation and a fixed codebook excitation, or the first historical linear prediction excitation may be an adaptive codebook excitation. Alternatively, the first historical linear prediction excitation may be another type of codebook excitation.

It is understood that the functions of the various functional modules of the audio encoder 600 of this embodiment can be specifically implemented according to the methods in the above-mentioned method embodiments. The specific implementation process can be referred to the relevant description of the above-mentioned method embodiments and will not be repeated here. The audio encoder 600 can be any device that needs to collect, store, or transmit audio signals, such as a mobile phone, tablet computer, personal computer, laptop computer, etc.

Among them, for examples of values of various thresholds (such as the first threshold, the second threshold, etc.) and other parameters (such as N1, N11, N21, N2, etc.) involved in the embodiment of this device, please refer to the relevant value examples in the above method embodiment, and no further details will be given here.

It can be seen that in the technical solution of this embodiment, the audio encoder 600 first estimates the reference linear prediction efficiency of the current audio frame; determines the audio coding method that matches it through the estimated reference linear prediction efficiency of the above-mentioned current audio frame, and performs audio encoding on the above-mentioned current audio frame according to the determined audio coding method that matches it. Since the above-mentioned scheme does not need to perform the operation of fully encoding the current audio frame using each audio coding method required by the existing closed-loop selection mode in the process of determining the audio coding method, but instead determines the audio coding method to be selected through the reference linear prediction efficiency of the current audio frame, and the computational complexity of estimating the reference linear prediction efficiency of the current audio frame is usually much smaller than the computational complexity of fully encoding the current audio frame using each audio coding method, compared with the existing mechanism, the above-mentioned scheme of the embodiment of the present invention is conducive to reducing the computational complexity of audio coding, thereby reducing the overhead of audio coding.

An embodiment of the present invention further provides a computer storage medium, wherein the computer storage medium may store a program, and when the program is executed, the program includes part or all of the steps of any one of the audio encoding methods recorded in the above method embodiments.

It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program codes.

As described above, the above embodiments are merely used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (16)

1. An audio encoding method, characterized in that it comprises: When the current audio frame is a non-speech audio frame, estimate the reference linear prediction efficiency of the current audio frame; Determine an audio coding scheme that matches the reference linear prediction efficiency of the current audio frame; The current audio frame is encoded using an audio encoding method that matches the reference linear prediction efficiency of the current audio frame. The reference linear prediction efficiency includes at least one of the following linear prediction efficiencies: reference long-term linear prediction efficiency, reference short-term linear prediction efficiency, and reference comprehensive linear prediction efficiency. 2. The method according to claim 1, characterized in that, If the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then determining the audio coding scheme that matches the reference linear prediction efficiency of the current audio frame includes: If the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to the fifth threshold, then the audio coding method that matches the reference linear prediction efficiency of the current audio frame is determined to be the audio coding method based on linear prediction. And/or, If the reference short-time linear prediction efficiency of the current audio frame is less than the fifth threshold, then the audio coding method that matches the reference linear prediction efficiency of the current audio frame is determined to be a non-linear prediction-based audio coding method. 3. The method according to claim 1, characterized in that, If the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, then determining the audio coding method that matches the reference linear prediction efficiency of the current audio frame includes: determining the second linear prediction efficiency interval into which the reference short-time linear prediction efficiency of the current audio frame falls; and determining a second audio coding method that has a mapping relationship with the second linear prediction efficiency interval based on the mapping relationship between the linear prediction efficiency interval and audio coding methods based on linear prediction, wherein the second audio coding method is an audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the second audio coding method is an audio coding method based on linear prediction or an audio coding method not based on linear prediction. 4. The method according to any one of claims 1 to 3, characterized in that, The reference short-time linear prediction efficiency of the current audio frame is estimated as follows: the short-time linear prediction efficiency of the current audio frame is estimated, wherein the short-time linear prediction efficiency of the current audio frame is the reference short-time linear prediction efficiency of the current audio frame. or, The reference short-time linear prediction efficiency of the current audio frame is estimated as follows: the short-time linear prediction efficiency of the current audio frame is estimated; the linear prediction efficiency of N5 historical audio frames of the current audio frame is obtained; a fourth statistical value of the linear prediction efficiency of the N5 historical audio frames and the short-time linear prediction efficiency of the current audio frame is calculated, wherein N5 is a positive integer, and the fourth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the linear prediction efficiency of each of the N51 historical audio frames is at least one of the following linear prediction efficiencies: long-time linear prediction efficiency, short-time linear prediction efficiency, and combined linear prediction efficiency, and the N51 historical audio frames are a subset of the N5 historical audio frames; or, The reference short-time linear prediction efficiency of the current audio frame is estimated as follows: the short-time linear prediction efficiency of the current audio frame is estimated; the reference linear prediction efficiency of N6 historical audio frames of the current audio frame is obtained; a fifth statistical value of the reference linear prediction efficiency of the N6 historical audio frames and the short-time linear prediction efficiency of the current audio frame is calculated, wherein N6 is a positive integer, the fifth statistical value is the reference short-time linear prediction efficiency of the current audio frame, and the reference linear prediction efficiency of each of the N61 historical audio frames is at least one of the following linear prediction efficiencies: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency, and reference comprehensive linear prediction efficiency, wherein the N61 historical audio frames are a subset of the N6 historical audio frames; or, The reference short-time linear prediction efficiency of the current audio frame is estimated as follows: The short-time linear prediction efficiency of the current audio frame is estimated; the reference linear prediction efficiency of N8 historical audio frames of the current audio frame is obtained; the linear prediction efficiency of N7 historical audio frames of the current audio frame is obtained; a sixth statistical value is calculated of the linear prediction efficiency of the N7 historical audio frames, the reference linear prediction efficiency of the N8 historical audio frames, and the short-time linear prediction efficiency of the current audio frame, where N7 and N8 are positive integers, and the sixth statistical value is the reference short-time linear prediction efficiency of the current audio frame. The linear prediction efficiency of each historical audio frame in the N71 historical audio frames is at least one of the following linear prediction efficiencies: long-term linear prediction efficiency, short-term linear prediction efficiency, and combined linear prediction efficiency. The reference linear prediction efficiency of each historical audio frame in the N81 historical audio frames is at least one of the following linear prediction efficiencies: reference long-term linear prediction efficiency, reference short-term linear prediction efficiency, and reference combined linear prediction efficiency. The N71 historical audio frames are a subset of the N7 historical audio frames, and the N81 historical audio frames are a subset of the N8 historical audio frames. 5. The method according to claim 4, wherein estimating the short-time linear prediction efficiency of the current audio frame includes: obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame. 6. The method according to claim 5, characterized in that, obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame includes: Calculate the energy change rate before and after performing short-time linear prediction on the current audio frame, wherein the energy change rate is the short-time linear prediction efficiency of the current audio frame, or the short-time linear prediction efficiency of the current audio frame is obtained by transforming the energy change rate, wherein the energy of the current audio frame after performing short-time linear prediction is the energy of the linear prediction residual of the current audio frame. 7. The method according to claim 6, wherein the energy change rate of the current audio frame before and after short-time linear prediction is the ratio of the energy of the current audio frame before short-time linear prediction to the energy of the linear prediction residual of the current audio frame. 8. An audio encoder, characterized in that it comprises: An estimation unit is used to estimate the reference linear prediction efficiency of the current audio frame when the current audio frame is a non-speech audio frame. A determining unit is used to determine an audio coding scheme that matches the reference linear prediction efficiency of the current audio frame estimated by the estimation unit; The encoding unit is used to encode the current audio frame according to the audio encoding method determined by the determining unit that matches the reference linear prediction efficiency of the current audio frame; The reference linear prediction efficiency includes at least one of the following linear prediction efficiencies: reference long-term linear prediction efficiency, reference short-term linear prediction efficiency, and reference comprehensive linear prediction efficiency. 9. The audio encoder according to claim 8, wherein if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, the determining unit is specifically used for: If the reference short-time linear prediction efficiency of the current audio frame is greater than or equal to the fifth threshold, then the audio coding method that matches the reference linear prediction efficiency of the current audio frame is determined to be the audio coding method based on linear prediction. And/or, If the reference short-time linear prediction efficiency of the current audio frame is less than the fifth threshold, then the audio coding method that matches the reference linear prediction efficiency of the current audio frame is determined to be a non-linear prediction-based audio coding method. 10. The audio encoder according to claim 8, wherein if the reference linear prediction efficiency of the current audio frame includes the reference short-time linear prediction efficiency of the current audio frame, the determining unit is specifically configured to: determine the second linear prediction efficiency interval into which the reference short-time linear prediction efficiency of the current audio frame falls; and determine a second audio coding method that has a mapping relationship with the second linear prediction efficiency interval based on the mapping relationship between the linear prediction efficiency interval and the audio coding method based on linear prediction, wherein the second audio coding method is an audio coding method that matches the reference linear prediction efficiency of the current audio frame, and the second audio coding method is an audio coding method based on linear prediction or an audio coding method not based on linear prediction. 11. The audio encoder according to any one of claims 8 to 10, In estimating the reference short-time linear prediction efficiency of the current audio frame, the estimation unit is specifically configured to: estimate the short-time linear prediction efficiency of the current audio frame, wherein the short-time linear prediction efficiency of the current audio frame is the reference short-time linear prediction efficiency of the current audio frame. or, In estimating the reference short-time linear prediction efficiency of the current audio frame, the estimation unit is specifically configured to: estimate the short-time linear prediction efficiency of the current audio frame; obtain the linear prediction efficiency of N5 historical audio frames of the current audio frame; calculate a fourth statistical value of the linear prediction efficiency of the N5 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein N5 is a positive integer, the fourth statistical value is the reference short-time linear prediction efficiency of the current audio frame, and the linear prediction efficiency of each of the N51 historical audio frames is at least one of the following linear prediction efficiencies: long-time linear prediction efficiency, short-time linear prediction efficiency, and combined linear prediction efficiency, wherein the N51 historical audio frames are a subset of the N5 historical audio frames; or, In estimating the reference short-time linear prediction efficiency of the current audio frame, the estimation unit is specifically configured to: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of N6 historical audio frames of the current audio frame; calculate a fifth statistical value of the reference linear prediction efficiency of the N6 historical audio frames and the short-time linear prediction efficiency of the current audio frame, wherein N6 is a positive integer, the fifth statistical value is the reference short-time linear prediction efficiency of the current audio frame, wherein the reference linear prediction efficiency of each of the N61 historical audio frames is at least one of the following linear prediction efficiencies: reference long-time linear prediction efficiency, reference short-time linear prediction efficiency, and reference comprehensive linear prediction efficiency, wherein the N61 historical audio frames are a subset of the N6 historical audio frames; or, In estimating the reference short-time linear prediction efficiency of the current audio frame, the estimation unit is specifically configured to: estimate the short-time linear prediction efficiency of the current audio frame; obtain the reference linear prediction efficiency of N8 historical audio frames of the current audio frame; obtain the linear prediction efficiency of N7 historical audio frames of the current audio frame; and calculate a sixth statistical value of the linear prediction efficiency of the N7 historical audio frames, the reference linear prediction efficiency of the N8 historical audio frames, and the short-time linear prediction efficiency of the current audio frame, wherein N7 and N8 are positive integers, and the sixth statistical value is the reference short-time linear prediction efficiency of the current audio frame. The linear prediction efficiency is as follows: for each of the N71 historical audio frames, the linear prediction efficiency is at least one of the following linear prediction efficiencies: long-term linear prediction efficiency, short-term linear prediction efficiency, and combined linear prediction efficiency; for each of the N81 historical audio frames, the reference linear prediction efficiency is at least one of the following linear prediction efficiencies: reference long-term linear prediction efficiency, reference short-term linear prediction efficiency, and reference combined linear prediction efficiency. The N71 historical audio frames are a subset of the N7 historical audio frames, and the N81 historical audio frames are a subset of the N8 historical audio frames. 12. The audio encoder of claim 11, in the aspect of estimating the short-time linear prediction efficiency of the current audio frame, the estimation unit is specifically configured to: obtain the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame. 13. The audio encoder according to claim 12, wherein, in the aspect of obtaining the short-time linear prediction efficiency of the current audio frame based on the linear prediction residual of the current audio frame, the estimation unit is specifically configured to: calculate the energy change rate before and after the current audio frame undergoes short-time linear prediction, wherein the energy change rate is the short-time linear prediction efficiency of the current audio frame, or the short-time linear prediction efficiency of the current audio frame is obtained based on the energy change rate, wherein the energy of the current audio frame after undergoing short-time linear prediction is the energy of the linear prediction residual of the current audio frame. 14. The audio encoder according to claim 13, wherein the energy change rate of the current audio frame before and after short-time linear prediction is the ratio of the energy of the current audio frame before short-time linear prediction to the energy of the linear prediction residual of the current audio frame. 15. An audio encoder, characterized in that the audio encoder includes a processor and a memory, the processor executing the method according to any one of claims 1 to 7 by calling a program or instruction stored in the memory. 16. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, wherein the computer program, when executed by a computer device, is capable of implementing the method described in any one of claims 1 to 7.