WO2012044067A1 - Method and apparatus for decoding an audio signal using an adaptive codebook update - Google Patents

Abstract

The present invention relates to a method and apparatus for decoding an audio signal using an adaptive codebook update. According to one embodiment of the present invention, the method for decoding an audio signal comprises the following steps: taking, as an input, an N^th frame, which is a frame with a loss, and an N+1^th frame, which is a normal frame; determining whether or not to update an adaptive codebook of the last subframe of the N^th frame using the N^th frame and N+1^th frame; updating the adaptive codebook of the last subframe of the N^th frame using the N+1t^th frame; and synthesizing an audio signal using the N^th frame. According to the present invention, the adaptive codebook of the last subframe of the frame with a loss is updated using the frame data normally received after a loss of frame data upon the occurrence of a loss of frame data, thus more quickly returning to the state before the loss of frame data.

Description

Method and apparatus for decoding audio signal using adaptive codebook update

The present invention relates to a method and apparatus for decoding an audio signal, and more particularly, to a method and apparatus for decoding an audio signal using an adaptive codebook update.

In order to transmit a voice (audio) signal for a voice (audio) call to a communication network, an encoder for compressing an audio signal converted into a digital signal and a decoder for recovering an audio signal from encoded data are used. One of the most widely used audio codec (encoder and decoder) technologies is Code Excited Linear Prediction (CELP). The CELP codec represents an audio signal with a synthesis filter modeling saints and the input signal from the synthesis filter.

Representative CELP codecs include G.729 and Adaptive Multi-Rate (AMR) codecs. The encoders of these codecs extract the coefficients of the synthesis filter from the input signal of one frame corresponding to 10 or 20 msec, and divide the frame into subframes of 5 msec again, the pitch index and gain of the adaptive codebook, and the pulse of the fixed codebook. Find the index and gain. The decoder also generates an excitation signal using the pitch index and gain of the adaptive codebook and the pulse index and gain of the fixed codebook, and restores the audio signal by filtering the excitation signal with a synthesis filter.

In the process of transmitting the frame data output from the encoder, frame loss may occur according to the state of the communication network. In order to reduce the degradation of the synthesized signal due to such frame loss, a frame loss concealment algorithm is used. Most frame loss concealment algorithms use the normal frame data before frame loss occurs to recover the audio signal of the lost frame. However, errors due to loss affect not only lost frames but also frames that are normally received later. Thus, when decoding an audio signal, if a normal frame is received after a frame loss has occurred, a method is needed to return to the state before the loss occurs as soon as possible.

According to the present invention, when frame data loss occurs, the audio code can be returned to the state before the frame loss more quickly by updating the adaptive codebook of the last subframe of the lost frame using the frame data normally received after the frame data loss. It is an object of the present invention to provide a signal decoding method and apparatus.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention, which are not mentioned above, can be understood by the following description, and more clearly by the embodiments of the present invention. Also, it will be readily appreciated that the objects and advantages of the present invention may be realized by the means and combinations thereof indicated in the claims.

According to an aspect of the present invention, there is provided a method of decoding an audio signal, the method comprising: receiving an N + 1 th frame, which is a normal frame transmitted after an N th frame, which is lost frame data, an N th frame and an N + 1 th frame Determining whether to update the adaptive codebook of the last subframe of the Nth frame using the method, updating the adaptive codebook of the last subframe of the Nth frame using the N + 1th frame, and audio using the Nth frame. And synthesizing the signal.

The present invention also provides an apparatus for decoding an audio signal, comprising: an input unit for receiving an N + 1 th frame, which is a normal frame transmitted after an N th frame, which is lost frame data, and an N th using an N th frame and an N + 1 th frame A control unit determines whether to update the adaptive codebook of the last subframe of the frame, updates the adaptive codebook of the last subframe of the Nth frame using the N + 1th frame, and a decoding unit to synthesize an audio signal using the Nth frame. It is another feature to include.

According to the present invention as described above, when frame data loss occurs, after the frame data loss, the adaptive codebook of the last subframe of the lost frame is updated by using the frame data normally received, so that the frame data loss can be brought to a state before the frame loss more quickly. There is an effect to return.

1 is a configuration diagram showing a configuration of a CELP encoder.

2 is a configuration diagram showing a configuration of a CELP decoder.

3 is a frame sequence transmitted from an encoder to a decoder.

4 is a flowchart illustrating a process of restoring an excitation signal in units of subframes after restoring the synthesis filter coefficients of the N-th frame lost by the AMR-WB codec.

5 is a flowchart illustrating a method of decoding an audio signal according to an embodiment of the present invention.

6 is a block diagram of an apparatus for decoding an audio signal according to an embodiment of the present invention.

The above objects, features, and advantages will be described in detail with reference to the accompanying drawings, whereby those skilled in the art to which the present invention pertains may easily implement the technical idea of the present invention. In describing the present invention, when it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description will be omitted. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to indicate the same or similar components.

1 is a configuration diagram showing the configuration of a CELP encoder.

Preprocessing filter 102 scales the input signal and performs high pass filtering. In this case, the input signal may have a length of 10 msec or 20 msec, and is composed of a plurality of subframes. In this case, the subframe generally has a length of 5 msec.

The LPC obtaining unit 104 extracts a linear prediction coefficient (LPC) corresponding to the coefficient of the synthesis filter from the preprocessed input signal. The LPC obtainer 104 then quantizes the extracted LPC and interpolates with the LPC of the previous frame to obtain the synthesis filter coefficients of each subframe.

The pitch analyzer 106 analyzes the pitch of the input signal in units of subframes to obtain a pitch index and a gain of the adaptive codebook. The obtained pitch index is used to reproduce the adaptive codebook value in the adaptive codebook module 112. The fixed codebook search unit 108 retrieves the fixed codebook of the input signal on a subframe basis to obtain a pulse index and a gain of the fixed codebook. The obtained pulse index is used to reproduce the fixed codebook value in the fixed codebook module 110. The adaptive codebook gain and the fixed codebook gain are quantized by the gain quantization unit 122.

The output of the fixed codebook module 110 reproduced by the pulse index is multiplied by the quantized fixed codebook gain (114). The output of the adaptive codebook module 112 reproduced by the pitch index is then multiplied by the quantized adaptive codebook gain (116). The gain multiplied by the adaptive codebook value and the fixed codebook value are then added to generate an excitation signal.

The generated excitation signal is input to the synthesis filter 118. Subsequently, the error between the input signal preprocessed by the preprocessing unit 102 and the output signal of the synthesis filter 118 is filtered by the cognitive weighting filter 120 reflecting the human auditory characteristics, and the pitch at which the error signal is the smallest. The index, the quantized gain, and the pulse index and the quantized gain are obtained and transmitted to the parameter encoder 124. The parameter encoding unit 124 encodes the pitch index of the adaptive codebook, the pulse index of the fixed codebook, the output of the gain quantization unit 122 and the LPC parameters in a form suitable for transmission, and outputs frame data. The output frame data is transmitted to the decoder via a network or the like.

2 is a configuration diagram illustrating a configuration of a CELP decoder.

The decoder reconstructs the fixed codebook 202 and the adaptive codebook 204 through the pulse index and pitch index transmitted from the encoder, respectively. The output of the fixed codebook 202 is then multiplied by the fixed codebook gain (206), and the output of the adaptive codebook 204 is multiplied by the adaptive codebook gain (208). The gain multiplied by the adaptive codebook value and the fixed codebook value are added to recover the excitation signal. The reconstructed excitation signal is filtered by a synthesis filter 210 consisting of coefficients produced by interpolating the LPC coefficients transmitted from the encoder. The output of the synthesis filter 205 is post-processed by the post processor 212 to restore the audio signal.

Meanwhile, in the process of transmitting the frame data output through the encoder of FIG. 1 to the decoder of FIG. 2, loss of the frame data may occur according to the state of the network. This frame data loss results in degradation of the synthesized audio signal at the decoder. To reduce the degradation of these audio signals, most codecs have a built-in frame loss concealment algorithm.

In the AMR-WB (Adaptive MultiRate-WideBand) codec, when a frame loss occurs, the data of the previous frame is scaled to a predetermined value according to the continuity of the frame loss and used as the lost frame data.

For example, as shown in FIG. 3, the N-th frame data is normally received, but when the N-th frame data is lost during transmission, the AMR-WB decoder first loses the data using the synthesis filter coefficients of the N-th frame. Restore the synthesis filter coefficients of the Nth frame. The pulse of the fixed codebook is recovered using a random function, and the scaled value is used as the fixed codebook gain after median filtering the fixed codebook gain of the previous subframe. In addition, the pitch index of the adaptive codebook is reconstructed using the pitch index of the last subframe of the N-1th frame or the pitch indexes of the previous subframe and a random value, and the scaled value after median filtering the gain of the previous subframe. Use as an adaptive codebook gain. The audio signal of the lost frame is restored using the restored frame data. Meanwhile, in FIG. 3, BFI (Bad Frame Indication) is information indicating whether a corresponding frame is a lost frame or a normal frame. When BFI is 0, the corresponding frame is a normal frame, and when 1, the corresponding frame is a lost frame.

4 is a flowchart illustrating a process of restoring an excitation signal in units of subframes after restoring the synthesis filter coefficients of the Nth frame lost in the AMR-WB codec.

Referring to FIG. 4, it is first determined whether the corresponding frame is terminated (402). Then, it is determined whether the frame is a lost frame, that is, whether BFI is 1 (404).

If no frame loss occurs, the pitch index of the adaptive codebook is decoded from the received frame data (406) and the adaptive codebook is decoded using this (408). The fixed codebook is decoded using the pulse index (410), and the gain 412 of each codebook is decoded and multiplied by each codebook. Next, two codebooks are added to synthesize the excitation signal (414), and the synthesized excitation signal is filtered by the synthesis filter (416).

If frame loss occurs in step 404 (that is, BFI is 1), the pitch index of the adaptive codebook is restored using the pitch of the previous subframe (418), and the adaptive codebook is restored using the recovered pitch index. (420). The fixed codebook is recovered using a random function (422), and the adaptive codebook gain and the fixed codebook gain of the previous subframe are respectively median filtered and scaled to recover the adaptive codebook gain and the fixed codebook gain of the lost frame (424). ). The excitation signal is synthesized using the reconstructed adaptive codebook and the fixed codebook parameter (414), and the synthesized excitation signal is filtered by the synthesis filter (416).

As described above, the process of restoring the pitch index and the gain of the adaptive codebook and the pulse index and the gain of the fixed codebook when the frame loss occurs is different from when the frame loss does not occur. After that, the process of synthesizing the excitation signal and filtering it with a synthesis filter is the same regardless of whether or not the frame is lost.

When frame data loss occurs, the effects of frame loss affect not only the lost frame but also the normal frame received later. Therefore, in order to reduce the quality deterioration due to the frame loss, it is important to restore the audio signal of the frame in which the loss occurs, but it is also important to quickly restore the state before the loss occurs when the frame data is normally received later.

The present invention updates the adaptive codebook of the last subframe of the lost frame by using the pitch information of the first normal frame data received after the loss, in order to quickly escape the effects of the loss when the frame data is lost. .

5 is a flowchart illustrating a method of decoding an audio signal according to an embodiment of the present invention. In this embodiment, when the N-th frame is lost, and when the N + 1-th frame is a normal frame, the adaptive codebook of the last subframe of the N-th frame restored by the frame loss concealment algorithm is used. Re-update using to restore the quality degradation of the audio signal due to the loss of the N-th frame faster than when decoding the N + 1-th frame data.

Referring to FIG. 5, first, a first subframe pitch T0 and a second subframe pitch T0_2 of an N + 1th frame are decoded (504).

Next, whether the N + 1th frame is the first normal frame received after the lost frame (ie, prev_BFI == 1 & BFI == 0), and whether the current subframe is the first subframe of the N + 1th frame (I_subfr == 0) and whether the pitch (PrevT0) of the last subframe of the lost Nth frame and the pitch (T0) of the first subframe of the N + 1th frame are different (prev_T0! = T0) (506). If this condition is not met, the general decoding procedure after step 518 is performed.

If the conditions of step 506 are all satisfied, it is determined whether the absolute value of the pitch difference (T0-T0_2) of the first subframe and the second subframe of the N + 1th frame is smaller than the predetermined reference value (x). Determine (508). If this condition is not met, the general decoding procedure after step 518 is performed.

If the condition of step 508 is also satisfied, as follows, before synthesizing the first subframe excitation signal of the N + 1 th frame, using the pitch of the first subframe of the N + 1 th frame, the last sub of the N th frame Update the adaptive codebook of the frame. That is, the adaptive codebook of the last subframe of the Nth frame is recovered (510), and a random function of the last subframe of the Nth frame is generated (512). The excitation signal of the last subframe of the Nth frame is recovered (514), and the adaptive codebook of the last subframe of the Nth frame is updated (516).

Thereafter, the excitation signal synthesis and the filtering by the synthesis filter of the N + 1 th frame are performed. That is, decoding 518 of the adaptive codebook and decoding 520 of the fixed codebook are performed, and the gain of the adaptive codebook and the gain of the fixed codebook are decoded (522). The excitation signal of the frame is synthesized using the decoded adaptive codebook and its gain, the fixed codebook and the gain (524), and the synthesized excitation signal is filtered by the synthesis filter (525).

6 is a block diagram of an apparatus for decoding an audio signal according to an embodiment of the present invention.

An apparatus for decoding an audio signal 602 according to an embodiment of the present invention includes an input unit 604, a control unit 606, and a decoding unit 608.

The input unit 604 receives frame data of an encoded audio signal through an encoding device. As described above, frame loss may occur while the frame data is transmitted to the decoding apparatus. In an embodiment of the present invention, the input unit 604 receives an N + 1 th frame, which is a normal frame transmitted after the N th frame, which is a lost frame.

The controller 606 determines whether to update the adaptive codebook of the last subframe of the Nth frame by using the Nth frame and the N + 1th frame. According to this determination, the N + 1th frame is used to update the adaptive codebook of the last subframe of the Nth frame.

In one embodiment of the present invention, the controller 606 first decodes the first subframe pitch T0 and the second subframe pitch T0_2 of the N + 1 th frame.

Next, the controller 606 determines whether it is the first normal frame received after the lost frame of the N + 1th frame (ie, prev_BFI == 1 & BFI == 0), and if the current subframe is the N + 1th frame. Whether it is the first subframe (i_subfr == 0) and whether the pitch (PrevT0) of the last subframe of the lost Nth frame and the pitch (T0) of the first subframe of the N + 1th frame are different (prev_T0! = To determine T0). If this condition is not satisfied, the controller 606 performs a general decoding procedure.

If all of the above conditions are satisfied, the controller 606 determines that the absolute value of the pitch difference (T0-T0_2) between the first subframe and the second subframe of the N + 1th frame is greater than the predetermined reference value (x). Determine if it is small. If this condition is not satisfied, the controller 606 performs a general decoding procedure.

If the above condition is also satisfied, the controller 606 uses the pitch of the first subframe of the N + 1th frame to synthesize the first subframe excitation signal of the N + 1th frame before the last subframe of the Nth frame. Update the adaptive codebook of. That is, the controller 606 restores the adaptive codebook of the last subframe of the Nth frame and generates a random function of the last subframe of the Nth frame. The control unit 606 restores the excitation signal of the last subframe of the Nth frame, and updates the adaptive codebook of the last subframe of the Nth frame.

Thereafter, the controller 606 performs filtering by the excitation signal synthesis and the synthesis filter of the N + 1th frame. That is, the controller 606 performs decoding of the adaptive codebook and decoding of the fixed codebook, and decodes the gain of the adaptive codebook and the gain of the fixed codebook. Using the decoded adaptive codebook and its gain, the fixed codebook and its gain, the control unit 606 synthesizes the excitation signal of the corresponding frame, and the synthesized excitation signal is filtered by the synthesis filter.

According to the present invention as described above, when frame data loss occurs, after the frame data loss, the adaptive codebook of the last subframe of the lost frame is updated by using the frame data normally received, so that the frame data loss can be brought to a state before the frame loss more quickly. There is an effect to return.

In addition, according to the present invention, when the normal frame is lost, particularly in a transition period from unvoiced sound to voiced sound or a pitch change period, the frame loss can be quickly recovered to reduce the degradation of the synthesized signal due to the frame loss.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited by.

Claims (6)

In the method for decoding an audio signal, Receiving an N + 1 th frame, which is a normal frame transmitted after the N th frame, which is lost frame data; Determining whether to update an adaptive codebook of the last subframe of the Nth frame using the Nth frame and the N + 1th frame; Updating the adaptive codebook of the last subframe of the Nth frame using the N + 1th frame; And Synthesizing an audio signal using the Nth frame Decoding method of an audio signal comprising. The method of claim 1, Determining whether to update the adaptive codebook of the last subframe of the Nth frame A first determining step of determining whether a pitch of a last subframe of the Nth frame and a pitch of the first subframe of the N + 1th frame are different from each other; A second determination step of determining whether a difference between a pitch of a first subframe of the N + 1th frame and a pitch of a second subframe of the N + 1th frame is smaller than a predetermined reference value; And Determining that the adaptive codebook of the last subframe of the Nth frame is updated if both the results of the first determination step and the second determination step are positive. Decoding method of an audio signal comprising. The method of claim 1, Updating the adaptive codebook of the last subframe of the N-th frame Updating the adaptive codebook of the last subframe of the Nth frame using the pitch of the first subframe of the N + 1th frame. Decoding method of an audio signal comprising. An apparatus for decoding an audio signal, An input unit for receiving an N + 1 th frame, which is a normal frame transmitted after the N th frame, which is lost frame data; The Nth frame and the N + 1th frame are used to determine whether to update the adaptive codebook of the last subframe of the Nth frame, and the N + 1th frame is used to adapt the last subframe of the Nth frame. A controller for updating the codebook; And A decoding unit for synthesizing an audio signal using the Nth frame Apparatus for decoding an audio signal comprising. The method of claim 1, The control unit Whether the pitch of the last subframe of the Nth frame and the pitch of the first subframe of the N + 1th frame are different from each other, the pitch of the first subframe of the N + 1th frame, and the N + 1th frame It is determined whether the difference in pitch of the second subframe is smaller than a predetermined reference value, and if the determination result is all positive, the adaptive codebook of the last subframe of the Nth frame is updated. A device for decoding an audio signal. The method of claim 1, The control unit Updating the adaptive codebook of the last subframe of the Nth frame using the pitch of the first subframe of the N + 1th frame Method of decoding an audio signal.

Images