JP2002116799A - Audio signal encoding device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform fading processing when an audio signal is encoded so that noise such as a click sound is not generated. SOLUTION: When a quantization and encoding part 13 quantizes an audio spectrum signal outputted from a time-frequency conversion part 11 by using an auditory parameter SMR(Signal-to-Mask ratio) outputted from an auditory model part 12 and outputs the quantized signal, the quantization and encoding part 13 quantizes the audio spectrum signal by using a 1st variable for controlling the encoding quantity of the quantized signal and a 2nd variable for controlling the quantization distortion of the quantized signal and then vary the 1st variable with a parameter fp for fading.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio signal encoding apparatus for performing encoding after converting an input audio signal into a frequency domain.

[0002]

2. Description of the Related Art Conventionally, an audio signal encoding method for encoding an input audio signal by high efficiency encoding includes, for example, an adaptive spectral perceptual encoding method (ASPEC: Adaptive Spectral Perceptual).
Entropy Coding) and MPEG (Moving Picture Ex)
pert Group) 1 audio layer 3 or MPEG2
Audio AAC (Advanced Audio Coding) and the like.

A conventional audio signal encoding apparatus to which the above-described audio signal encoding method is applied will be described with reference to FIGS. 6 and 7.

FIG. 6 is a block diagram showing a conventional audio signal encoding apparatus, FIG. 7 is a flow chart showing a double loop operation at the time of quantization in the conventional audio signal encoding apparatus, and FIG. It is the schematic diagram which showed the correspondence with the band sfb with respect to a spectrum signal.

The conventional audio signal encoding device 10B shown in FIG. 6 comprises a time-frequency converter 11, an auditory model 12, a quantization encoder 13, and a bit stream generator 14. .

First, a digital audio signal (hereinafter referred to as an audio PCM signal) on a time axis is converted into a time-frequency signal by a time-frequency converter 1 in a frame unit having a substantially constant processing period.
1 and the auditory model unit 12 are input in parallel.

In the above-mentioned time-frequency converter 11, the input audio / PCM (Pulse-Code Modulation) is input.
FFT (Fast Fourier Transform) processing and MDCT (Modified Discrete Cosine Transform)
) The conversion from the time axis to the frequency axis is performed using processing or the like, and the converted audio spectrum signal is sent to the quantization encoding unit 13.

On the other hand, in the above-mentioned auditory model unit 12, a signal-to-masking ratio SMR, which is an auditory parameter obtained by calculating a masking level based on human psychological psychology, is used.
(Signal-to-Mask-ratio) is sent to the quantization encoding unit 13.

Next, in the above-mentioned quantization encoding unit 13,
The audio spectrum signal output from the time-frequency converter 11 is quantized for each band. Here, when performing quantization on the audio spectrum signal, a double loop is formed by an iteration loop (repetition loop) described later for nonlinear quantization and Huffman coding, and the code at the time of quantization is formed. The amount and the quantization distortion are controlled.

At this time, in the double loop of the iteration loop, the inner loop is included in the outer loop, and the number of bits used for the audio / spectrum signal from the time-frequency converter 11 is predetermined in the inner loop. A quantized signal is obtained while performing control so as to be within the range of the number of bits. Thereafter, the quantization distortion of the quantized signal is calculated in the outer loop, and the quantization distortion is controlled to be equal to or less than the allowable noise level based on the signal-to-masking ratio SMR output from the auditory model unit 12,
When the quantization distortion falls below the allowable noise level, the quantization signal is sent to the bit stream generator 14.

More specifically, as shown in FIG. 7 (a), when the iteration loop ITR is started in step 1, the process immediately shifts to the inner loop IR in the outer loop OR in step 2, and FIG. The inner loop IR is started as shown in b).

When the inner loop IR is started, the audio spectrum signal from the time-frequency converter 11 is quantized for each band in step 2a to obtain a quantized signal. Next, in step 2b, the number of bits used is calculated for the quantized signal by Huffman coding. Next, in step 2c, it is determined whether or not the number of bits used of the quantized signal is within a predetermined number of bits. If the number of bits used of the quantized signal does not fall within the predetermined number of bits (NO) in step 2c, global_gain is adjusted to increase in step 2d, so that all the bands in the quantized signal are adjusted. , The level is uniformly changed, and thereafter, the process returns to the above-described step 2a, and performs the above-described steps 2a to 2c again.
In step 2c, the process is repeated until the number of bits used of the quantized signal falls within the predetermined number of bits. On the other hand, in step 2c, if the number of bits used of the quantized signal is within the predetermined number of bits (YES), global_gain is determined. Thereafter, in step 2e, the inner loop IR ends, and the process returns to the outer loop OR.

Note that the predetermined number of bits in step 2c is
It means the number of bits that can be used in one audio frame obtained from a preset bit rate.

At this time, when quantizing the audio spectrum signal from the time-frequency converter 11 for each band in step 2a in the inner loop IR,
Quantization is performed according to a quantization equation shown in [Equation 1] below.

[0015]

(Equation 1) In the above [Equation 1], quant (k) indicates a quantization value for an index k of a quantization signal in a frame. Also, gl shown on the denominator side in [Equation 1]
obal_gain is the first for varying the level for all audio spectrum signals in the frame.
, And the first variable is an integer value. Also,
| Mdct_line shown on the molecule side in [Equation 1]
(K) | indicates the absolute level of the audio spectrum signal with respect to the index k. Furthermore, [Equation 1]
Scalefactor (sfb) shown on the molecular side in
Is a second variable for controlling the quantization distortion by varying the level of the audio spectrum signal in band sfb units. At this time, the correspondence between the audio spectrum signal and the band sfb is schematically shown in FIG.
It is like.

Next, the inner loop I in step 2
When R is completed, the outer loop O shown in FIG.
The process proceeds to step 3 in R, and in step 3, inverse quantization is performed based on the quantization result obtained by the inner loop IR, and quantization distortion is calculated for each band. Here, when performing the inverse quantization, the inverse quantization is performed based on the inverse quantization formula shown in [Equation 2] below.

[0017]

(Equation 2) Next, in step 4, the quantization distortion calculated for each band is used as the signal-to-masking ratio SMR of the auditory model unit 12.
Judge whether it is within the allowable distortion obtained from
If the quantization distortion is not within the allowable distortion (NO), the scalefactor of that band is determined in step 5.
r (sfb) is adjusted to increase. In this case, since there is one or more bands in which the quantization distortion is not within the allowable distortion, the process returns to the above-described step 2 and performs the above-described steps 2 to 4 again. Repeat until it falls within the allowable distortion. On the other hand, if the quantization distortion falls within the allowable distortion in step 4 (YES), the scalefactor
(Sfb) is determined. Thereafter, in step 6, the quantized signal and the first variable global_gain
And the second variable, scalefactor (sf
b) is output to the bit stream generator 14, and the iteration loop ITR ends in step S7.

Returning to FIG. 6, in the bit stream generator 14, the quantized signal output from the quantization encoder 13 and
global_gain and scalefactor
(Sfb) is multiplexed and a bit stream is output, thereby completing the encoding of the audio signal.

[0019]

By the way, when encoding the audio / PCM signal (content) by applying the above-mentioned conventional audio signal encoding apparatus, when the program is switched or the music is switched, the content side fades. It is desirable to perform encoding in the state where is applied, but when the fade processing on the content side is not applied (when the connection part is discontinuous), level fluctuation occurs when encoding / decoding This causes a problem such as generation of an abnormal sound such as a click sound.

[0020]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and an audio signal encoding apparatus for encoding an audio signal converts the input audio signal from a time axis to a frequency axis. Audio
A time-frequency conversion unit that outputs a spectrum signal, an auditory model unit that calculates and outputs auditory parameters based on human auditory characteristics from the input audio signal, and an input fade that is used to perform a fade process on the audio signal. A fade parameter input unit for checking the parameters for the fade and outputting the parameters for the fade, and the audio spectrum signal output from the time-frequency conversion unit using the auditory parameters output from the auditory model unit. A quantization encoding unit that performs quantization and outputs a quantized signal, and a bit stream conversion unit that converts the quantized signal output from the quantization encoding unit into a bit stream and outputs the bit stream. The quantization coding unit includes a first variable for controlling a code amount of the quantization signal; And a second variable for controlling a quantization distortion of the child signal. After quantizing the audio spectrum signal using the second variable, the first variable is changed by the fade parameter. A signal encoding device is provided.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an audio signal encoding apparatus according to the present invention will be described below in detail with reference to FIGS.

FIG. 1 is a block diagram showing an audio signal encoding apparatus according to the present invention. FIG. 2 is a flowchart showing the operation of a double loop during quantization in the audio signal encoding apparatus according to the present invention. FIG. 3 is a flowchart showing the operation of the fade processing shown in FIG. 2, and FIG. 4 shows an audio signal decoding apparatus for decoding a bit stream output from the audio signal encoding apparatus according to the present invention. FIG. 5A is a waveform diagram illustrating a decoded output example when a fade parameter is not used, and FIG. 5B is a waveform diagram illustrating a decoded output example when a fade parameter is used. It is.

For convenience of explanation, the same reference numerals are given to the same constituent members as those shown in the conventional example, and the description is appropriately given, and new reference numerals are given to constituent members different from the conventional example. In addition, this embodiment will be described focusing on points different from the conventional example.

The audio signal encoding apparatus according to the present invention is characterized by performing a fade process when encoding an audio signal so as not to generate an abnormal sound such as a click sound. is there.

The audio signal encoding apparatus 10A according to the present invention shown in FIG. 1 includes a time-frequency conversion unit 11, an auditory model unit 12, a quantization encoding unit 13, a bit stream conversion unit 14, It comprises an added fade parameter input unit 15.

The input audio / PCM signal is sent to the time-frequency conversion unit 11 and the auditory model unit 12 on a frame-by-frame basis with a substantially constant processing period. Then, the converted audio / spectrum signal is sent to the quantization encoding unit 13. On the other hand, the signal-to-masking ratio SMR, which is an auditory parameter obtained by calculating the masking level based on the auditory psychology of the human in the auditory model unit 12, is sent to the quantization encoding unit 13.

Next, a fade parameter input unit 15 which is a main part of the present invention checks a fade parameter input from outside to input the audio / PCM in order to perform a fade process on a frame basis. , And the fade parameter fp to the quantization encoding unit 13.

Here, the fade parameter fp supplied to the quantization encoder 13 is 0 or a natural number (1, 2,..., N) set for the frame. Then, when the frame range to be subjected to the fade processing is set for the audio / PCM signal, the fade parameter fp is set to 0 or an appropriate numerical value by a natural number according to the frame state in which the fade processing is to be performed for each frame. The amplitude level at the time of fading can be changed by the quantization encoding unit 13.

Next, in the quantization encoding unit 13, when quantizing the audio spectrum signal output from the time-frequency conversion unit 11 for each band, the quantization is performed in the same manner as described in the conventional example. End the loop ITR,
A fade process is performed on the quantized signal at the stage where the iteration loop ITR has been completed, based on the fade parameter fp sent from the fade parameter input unit 15.

That is, in the above-mentioned quantization encoding unit 13,
As shown in FIGS. 2A and 2B, a double loop by the same iteration loop ITR as described in the conventional example is performed.
If only the results of the inner loop IR and the outer loop OR in R are described, when each step by the inner loop IR is completed, the number of bits used of the quantized signal is within a predetermined number of bits, and the quantized signal is The first variable global_gain for changing the level for all the bands in the band is determined.

After that, when each step by the outer loop OR is completed, the quantization distortion becomes equal to or less than the allowable noise level in band units, and scalefactor (sf) which is a second variable for controlling the quantization distortion is used.
b) is also determined.

Then, in step 6 'in the outer loop OR, the quantized signal and glo which is the first variable
bal_gain and the second variable, scalef
The actor (sfb) is output to the following fade processing side in the quantization encoding unit 13 without being output to the bit stream conversion unit 14 as described in the conventional example.

After that, when the iteration loop ITR is completed in step 7, the process enters a fade process F in step 8. In the fade process F, as shown in FIG. 3, in step 8a, the global_gain determined at the end of the iteration loop ITR and the supplied (global_gain) are considered in consideration of the fade parameter fp. ｛(Global_g
ain) -fp}.

Thereafter, in step 8b, the quantized signal and the replaced 置 き 換 え (globa) which is the first variable
l_gain) -fp} and the second variable sca
The factor (sfb) is sent from the quantization encoding unit 13 to the bit stream generation unit 14.

Returning to FIG. 1, in the bit stream generator 14, the quantized signal output from the quantization encoder 13 and
Replaced {(global_gain) -fp}
And scalefactor (sfb) are multiplexed to output a bit stream, thereby completing the encoding of the audio signal.

Next, when the audio signal encoding apparatus 10A according to the present invention performs a fade process based on the fade parameter fp and decodes the obtained bit stream, the audio signal decoding shown in FIG. Audio signal decoding device 20
Will be described briefly.

In the audio signal decoding apparatus 20 described above, the bit stream output from the bit stream forming section 14 of the audio signal coding apparatus 10A according to the present invention is input to the bit stream decomposing section 21, From the bit stream multiplexed by the unit 21, a signal corresponding to a quantized signal and ｛(globa)
l_gain) -fp}, and scale
The signal is decomposed into a signal corresponding to factor (sfb) and sent to the inverse quantization decoding unit 22.
Then, inverse quantization is performed on each signal and decoding is performed, and each decoded signal is obtained as a signal on the frequency axis.

Here, the quantized signal is global_g before being replaced using the fade parameter fp.
This is a result of quantization using “ain” and [Equation 1] described above. Normally, the inverse quantization of the quantized signal according to [Equation 2] described above is performed by replacing the first variable, ie, the replaced ｛(global_ga).
in) -fp}, the following [Equation 3] is obtained.
This is equivalent to the case where inverse quantization is performed according to the inverse quantization formula shown in (1), and the amplitude level after decoding can be attenuated.

[0039]

(Equation 3) The fade parameter f used here is
As for p, a numerical value is set in accordance with the frame condition to be faded as described above. For example, if the dB value is calculated when the fade parameter fp = 1, dB
The value is calculated by 20 × log ₁₀ {[Equation 2] / [Equation 3]}, that is, 20 × log 10
₁₀ ｛2 ^{(−fp / 4)} ｝ dB, so that f
By substituting p = 1, 20 × log ₁₀ ｛2
⁽ −1/4 ⁾ ΔdB = −1.505 dB

Further, in order to continuously attenuate the amplitude level after decoding, ｛(global_gai) in [Equation 3] is used.
n) It is necessary to increase the value of the fade parameter fp in -fp} in frame units. For example,
There is a method of increasing the fade parameter fp by 1, 2, 3,... One by one, or a method of increasing the fade parameter fp by 1, 3, 5,. For this reason, when performing the fade processing F in the quantization encoding unit 13, it is necessary to input the value of the fade parameter fp corresponding to each of the above methods to the fade parameter input unit 15.

Thereafter, the signals inversely decoded by the inverse quantization decoding unit 22 are sent to a frequency-time conversion unit 23, where the conversion from the frequency axis to the time axis is performed. Then, the signal is returned to the original audio / PCM signal and output from the frequency / time conversion unit 23.

Next, referring to FIGS. 5 (a) and 5 (b), a comparison is made between a decoded output example when the fade parameter fp is not used and a decoded output example when the fade parameter fp is used. I will explain.

First, as shown in FIG. 5A, an example of a decoded output example in which a bit stream encoded without using the fade parameter fp is decoded by the audio signal decoding device 20 will be described. , Frame 1 and frame 2 are not set with the fade parameter fp.
When a long window, which is a type of window function, is opened and frame 1 and frame 2 are connected to each other, no fade processing is performed on the connected portion as in the addition result shown in FIG. Is discontinuous, the level of the connection portion is not attenuated, so that the sound becomes hard to hear.

On the other hand, as shown in FIG. 5B, an example of a decoded output example in which a bit stream encoded using the fade parameter fp is decoded by the audio signal decoding device 20 will be described. Since the fade parameter fp is not set for the frame 1 but the fade parameter fp is set for the frame 2, the long window is opened for the frames 1 and 2 in this state. When the frame 1 and the frame 2 are connected in this way, the level of the connected portion is attenuated because the connected portion is subjected to fade processing as shown in the addition result shown in FIG. .

[0045]

According to the audio signal encoding apparatus according to the present invention described in detail above, in particular, a time-frequency converter for converting an input audio signal from a time axis to a frequency axis and outputting an audio spectrum signal; An auditory model unit that calculates and outputs auditory parameters based on human auditory characteristics from the input audio signal, and a fade that outputs the fader parameter by checking the input fader parameter to fade the audio signal A parameter input unit, and a quantization encoding unit that performs quantization on the audio spectrum signal output from the time-frequency conversion unit using the auditory parameters output from the auditory model unit and outputs a quantized signal. And a bit for converting the quantized signal output from the quantization encoding unit into a bit stream and outputting the bit stream. In the case of including the streaming unit, the quantization encoding unit includes a first variable for controlling the code amount of the quantized signal and a second variable for controlling the quantization distortion of the quantized signal. After quantizing the audio spectrum signal by using the above, by changing the first variable by the fade parameter, the audio signal is encoded so as not to generate an abnormal sound such as a click sound. , A fade process can be performed.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an audio signal encoding device according to the present invention.

FIG. 2 is a flowchart showing an operation of a double loop at the time of quantization in the audio signal encoding apparatus according to the present invention.

FIG. 3 is a flowchart showing an operation of the fade processing shown in FIG. 2;

FIG. 4 is a block diagram showing an audio signal decoding device for decoding a bit stream output from the audio signal encoding device according to the present invention.

FIG. 5A is a waveform diagram illustrating a decoded output example when a fade parameter is not used, and FIG. 5B is a waveform diagram illustrating a decoded output example when a fade parameter is used. is there.

FIG. 6 is a block diagram showing a conventional audio signal encoding device.

FIG. 7 is a flowchart showing an operation of a double loop at the time of quantization in a conventional audio signal encoding device.

FIG. 8 shows a band s for an audio spectrum signal.
It is the schematic diagram which showed the correspondence with fb.

[Explanation of symbols]

 10A: audio signal encoding apparatus according to the present invention, 11: time-frequency conversion unit, 12: auditory model unit, 13: quantization encoding unit, 14: bit stream generation unit, 15: parameter input unit for fade, F: Fade processing, fp: Fade parameters.

Claims (1)

[Claims] 1. An audio signal encoding apparatus for encoding an audio signal, comprising: a time-frequency converting unit for converting the input audio signal from a time axis to a frequency axis to output an audio spectrum signal; A hearing model unit for calculating and outputting a hearing parameter based on a human hearing characteristic from a signal, and a fade for outputting a fade parameter after confirming an input fade parameter in order to perform a fade process on the audio signal. A parameter input unit, and a quantization code that performs quantization on the audio spectrum signal output from the time-frequency conversion unit using the auditory parameter output from the auditory model unit and outputs a quantized signal. A quantizing unit, and the quantized signal output from the quantization encoding unit. A bit stream conversion unit that converts the stream into a stream and outputs the stream. The quantization coding unit includes a first variable for controlling a code amount of the quantization signal, and a quantization of the quantization signal. An audio signal encoding apparatus, comprising: quantizing the audio spectrum signal using a second variable for controlling distortion, and changing the first variable according to the fade parameter.