JP4305868B2 - Speech encoding method and speech decoding method - Google Patents

Description

The present invention relates to a speech encoding method and speech decoding method for compressing a multi-channel speech signal with a variable length.

  As a method of compressing an audio signal with a variable length, the present inventor, in the previous application (Japanese Patent Application No. 9-289159), uses a plurality of predictors having different characteristics in the time domain for a single channel original digital audio signal. Prediction that calculates multiple linear prediction values of current signal from past signal, calculates prediction residual for each predictor from original digital audio signal and multiple linear prediction values, and selects the minimum value of prediction residual An encoding method is proposed.

  In the above method, a certain degree of compression effect can be obtained when the original digital audio signal has a sampling frequency = 96 kHz and the number of quantization bits = 20 bits. However, in recent DVD audio discs, the sampling frequency (twice this ( = 192 kHz) is used, and the number of quantization bits tends to be 24. Therefore, it is necessary to improve the compression rate. In addition, the sampling frequency and the number of quantization bits in multichannel may be different for each channel.

  By the way, since a compression method such as the predictive coding method has a variable compression rate (VBR: variable bit rate), when a multi-channel audio signal is predictively encoded, the amount of data for each channel greatly changes over time. To do. When such data is transmitted, it is transmitted as a data stream instead of parallel for each channel.

  Therefore, in order to enable reproduction (presentation) of such a variable length data stream in synchronization with each channel on the reproduction side (decoding side), the data stream stored in the input buffer is read and output to the decoder. Therefore, it is necessary to manage the decoding time indicating the timing for the output and the reproduction time indicating the timing for reading the decoded data accumulated in the output buffer and outputting (presenting) the data to a speaker or the like. Further, the playback side must manage the time for searching and playing back such a variable length data stream.

  Accordingly, an object of the present invention is to provide a speech encoding method and speech decoding method capable of managing the processing time on the playback side when a multi-channel speech signal is encoded with a variable compression rate.

  In order to achieve the above object, the present invention comprises the following means 1) and 2).

1) in response to audio signals of a plurality of channels to the exact channel or audio signal inputted to each channel taken mutually correlated, with obtaining the top sampled values, characteristics in the time domain by a plurality of different linear prediction methods A step of selecting and predicting a linear prediction method that predicts a linear prediction value of a current signal from the past and minimizes a prediction residual obtained from the predicted linear prediction value and the speech signal. When,
Packing the prediction residual with the number of bits based on the number-of-bits information when packing the prediction encoded data including the linear prediction method, the prediction residual, and a predetermined head sample value of each selected channel ; When,
Generating decoding time stamp information indicating timing for reading compressed data in the input buffer on the decoding side according to the amount of the compressed data packed;
User data including a packet header including the decoding time stamp information, a compressed PCM private header including the UPC / EAN number and ISRC data of the voice signal, and a compressed PCM access unit including the compressed data. Formatting into packets;
A speech encoding method comprising:
2) A speech decoding method for decoding the original speech signals of the plurality of channels from the data encoded by the speech encoding method according to claim 1,
Decoding the UPC / EAN number and ISRC data located in the compressed PCM private header;
Decoding the stored subpackets based on the decoding time stamp information to separate compressed PCM access units;
The prediction residual of the compressed data in the separated compressed PCM access unit is decoded with the number of bits based on bit information, and the prediction value is calculated based on the decoded prediction residual, the head sample value, and the linear prediction method. Calculating for each channel ;
Restoring the original audio signals of the plurality of channels from the calculated predicted values;
A speech decoding method comprising:

  As described above, according to the present invention, the decoding time stamp information indicating the timing for reading the compressed data is included in the packet header, so that reproduction is performed when a multi-channel audio signal is encoded with a variable compression rate. The side can perform search playback.

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a first embodiment of a voice encoding apparatus to which the present invention is applied and a corresponding voice decoding apparatus. FIG.
FIG. 3 is an explanatory diagram showing a bit stream encoded by the encoding unit shown in FIGS. 1 and 2, FIG. 4 is an explanatory diagram showing a format of a DVD pack, and FIG.
Is an explanatory diagram showing the format of a DVD audio pack, FIG. 6 is a block diagram showing in detail the decoding unit of FIG. 1, FIG. 7 is a timing chart showing write / read timing of the input buffer of FIG. 6, and FIG. FIG. 9 is an explanatory diagram showing an access unit and a presentation unit.

Here, as the multi-channel method, for example, the following four methods are known.
(1) 4-channel system Like the Dolby Surround system, a total of 4 channels of 3 channels for the front L, C, and R + 1 channel for the rear S (2) 5 channels system Like no Dolby AC-3 system SW channel , Front L, C, R 3 channels + rear SL, SR 2 channels in total 5 channels (3) 6 channel system 6 channels (L, L, DTS (Digital Theater System) system and Dolby AC-3 system C, R, SW (Lfe), SL, SR)
(4) 8-channel system Like the SDDS (Sony Dynamic Digital Sound) system, a total of 8 channels including 6 channels of forward L, LC, C, RC, R, and SW + 2 channels of backward SL and SR The 6-channel (ch) mix-and-matrix circuit 1 'on the control side includes, as an example of a multi-channel signal, front left (Lf), center (C), front right (Rf), surround left (Ls), and surround right (Rs) And Lfe (Low Frequency Effect) 6ch PCM data is classified and converted into 2ch “1” and “2” related to the front group and 4ch “3” to “6” related to the other group according to the following equation (1): 2ch “1” and “2” are output to the first encoding unit 2′-1, and 4ch “3” to “6” are output to the second encoding unit 2′-2.

“1” = Lf + Rf
“2” = Lf−Rf
“3” = C− (Ls + Rs) / 2
“4” = Ls + Rs
“5” = Ls−Rs
“6” = Lfe−a × C
However, 0 ≦ a ≦ 1 (1)
As shown in detail in FIG. 2, the first and second encoding units 2′-1, 2′-2 constituting the encoding unit 2 ′ are respectively 2ch “1”, “2” and 4ch “3” to “3” to “3” to “3”. The PCM data of “6” is predictively encoded, and the predictive encoded data is transmitted to the decoding side via the recording medium 5 and the communication medium 6 in a bit stream as shown in FIG. On the decoding side, the first and second decoding units 3′-1 and 3′-2 constituting the decoding unit 3 ′ perform 2ch “1” and “2” respectively related to the front group as shown in detail in FIG. The predictive encoded data of 4ch “3” to “6” related to other groups is decoded into PCM data.

  Next, the original 6ch (Lf, C, Rf, Ls, Rs, Lfe) is restored by the mix & matrix circuit 4 ′ based on the equation (1), and the original 6ch and the coefficient mij (i = 1, 2, j = 1, 2 to 6), stereo 2ch data (L, R) is generated as in the following equation (2).

L = m11 · Lf + m12 · Rf + m13 · C
+ M14 · Ls + m15 · Rs + m16 · Lfe
R = m21 · Lf + m22 · Rf + m23 · C
+ M24 · Ls + m25 · Rs + m26 · Lfe (2)
The encoding units 2′-1 and 2′-2 will be described in detail with reference to FIG. The PCM data of each channel “1” to “6” is stored in one frame buffer 10 for each frame. The sample data of each channel “1” to “6” of one frame is applied to the prediction circuits 13D1, 13D2, and 15D1 to 15D4, respectively, and the head sample data of each frame of each channel “1” to “6” Is applied to the formatting circuit 19. Each of the prediction circuits 13D1, 13D2, 15D1 to 15D4 outputs a current signal from a past signal in the time domain to a PCM data of each channel “1” to “6” by a plurality of predictors (not shown) having different characteristics. A plurality of linear prediction values are calculated, and then a prediction residual for each predictor is calculated from the original PCM data and the plurality of linear prediction values. The subsequent buffer / selectors 14D1, 14D2, 16D1 to 16D4 temporarily store the prediction residuals calculated by the prediction circuits 13D1, 13D2, and 15D1 to 15D4, respectively, and select signals / DTS (decoding time stamps). The minimum value of the prediction residual is selected for each subframe designated by the generator 17.

  The selection signal / DTS generator 17 applies a bit number flag of the prediction residual to the packing circuit 18 and the formatting circuit 19, and also includes a predictor selection flag indicating a predictor having the smallest prediction residual, and an expression ( The correlation coefficient a in 1) and the DTS indicating the time for the decoding side to extract the stream data from the input buffer 22a (FIG. 6) are applied to the formatting circuit 19. The packing circuit 18 packs the prediction residuals for 6ch selected by the buffers / selectors 14D1, 14D2, 16D1 to 16D4 with the designated number of bits based on the bit number flag designated by the selection signal / DTS generator 17. . The PTS generator 17c generates a PTS (Presentation Time Stamp) indicating the time when the decoding side takes out the PCM data from the output buffer 110 (FIG. 6), and outputs the PTS (Presentation Time Stamp) to the formatting circuit 19.

The subsequent formatting circuit 19 formats the user data as shown in FIGS. The user data (subpacket) shown in FIG. 3 includes a variable rate bit stream (substream) BS0 including 2ch “1” and “2” predictive encoded data related to the forward group, and 4ch “3” to “3” related to other groups. It is composed of a variable rate bit stream (substream) BS1 including predictive encoded data of “6” and a bitstream header (restart header) provided before the substreams BS0 and BS1.
Also, one frame of substream BS0, BS1 is a frame header,
・ First sample data of one frame of each channel “1” to “6”,
A predictor selection flag for each subframe of each channel “1” to “6”;
A bit number flag for each subframe of each channel “1” to “6”;
-Predictive residual data string (number of variable bits) of each ch "1" to "6",
・ Ch “6” coefficient a
Are multiplexed. According to such predictive coding, when the original signal is, for example, sampling frequency = 96 kHz, quantization bit number = 24 bits, and 6 channels, a compression rate of 71% can be realized.

  When the variable rate bit stream data predictively encoded by the encoding units 2′-1, 2′-2 shown in FIG. 2 is recorded on a DVD audio disk as an example of a recording medium, the audio ( A) Packed in a pack. This pack consists of 2034 bytes of user data (A packet, V packet), 4 bytes of pack start information, 6 bytes of SCR (System Clock Reference) information, and 3 bytes of Mux rate ( rate) information and a 1-byte stuffing total 14-byte pack header are added (1 pack = total 2048 bytes). In this case, the time of the A pack in the same title can be managed by setting the SCR information as a time stamp as “1” in the first pack and continuing in the same title.

As shown in detail in FIG. 5, the compressed PCM A packet is composed of a 19 or 14 byte packet header, a compressed PCM private header, and 1 to 2011 byte audio data (compressed PCM) in the format shown in FIG. ing. The DTS and PTS are set in the packet header of FIG. 5 (specifically, the PTS is in the 10th to 14th bytes of the packet header and the DTS is in the 15th to 19th bytes). The compressed PCM private header is
A 1-byte substream ID,
2-byte UPC / EAN-ISRC (Universal Product Code / European Article Number-International Standard Recording Code) number and UPC / EAN-I
SRC data,
-1 byte private header length,
A 2-byte first access unit pointer;
8 bytes of audio data information (ADI)
・ With stuffing byte of 0-7 bytes,
It is made up of. Both the forward access unit search pointer for searching the access unit after 1 second in the ADI and the backward access unit search pointer for searching for the access unit before 1 second are both 1 byte (specifically, Is set to the 7th byte of the ADI with the forward access unit search pointer and the 8th byte with the backward access unit search pointer.

  Next, the decoding units 3'-1 and 3'-2 will be described with reference to FIG. The variable rate bit stream data BS0 and BS1 in the above format are separated by the deformatting circuit 21. The first sample data of one frame of each channel “1” to “6” and the predictor selection flag are respectively applied to the prediction circuits 24D1, 24D2, 23D1 to 23D4, and the number of bits of each channel “1” to “6”. The flag is applied to the unpacking circuit 22. The SCR, DTS, and prediction residual data string are applied to the input buffer 22a, and PTS is applied to the output buffer 110. Here, a plurality of predictors (not shown) in the prediction circuits 24D1, 24D2, 23D1 to 23D4 have the same characteristics as the plurality of predictors in the encoding-side prediction circuits 13D1, 13D2, and 15D1 to 15D4, respectively. Those having the same characteristics are selected by the predictor selection flag.

  The stream data (predicted residual data string) separated by the deformatting circuit 21 is captured and stored in the input buffer 22a for each access unit by the SCR as shown in FIG. Here, the data amount of one access unit is, for example, (1/96 kHz) seconds when fs = 96 kHz, but is variable length as shown in detail in FIGS. 8 and 9A. The stream data stored in the input buffer 22a is read out by the FIFO based on the DTS and applied to the unpacking circuit 22.

The unpacking circuit 22 separates the prediction residual data strings of the channels “1” to “6” based on the bit number flags and outputs them to the prediction circuits 24D1, 24D2, and 23D1 to 23D4, respectively. Each of the prediction circuits 24D1, 24D2, 23D1 to 23D4 uses the current prediction residual data of each channel “1” to “6” from the unpacking circuit 22 and a predictor selection flag among a plurality of internal predictors. The previous prediction value predicted by each selected one is added to calculate the current prediction value, and then the PCM data of each sample is calculated and stored in the output buffer 110 with reference to the first sample data of one frame. Is done. The PCM data stored in the output buffer 110 is read and output based on the PTS. Therefore, the variable-length access unit shown in FIG. 9A is expanded, and the fixed-length presentation unit shown in FIG. 9B is output.

  Here, when search reproduction is instructed via the operation unit 101, the control unit 100 places the forward access unit / search pointer indicating one second ahead placed in the ADI shown in FIG. 5 and the backward indicating one second later. The access unit is reproduced based on the access unit search pointer. This search pointer may be one second ahead and two seconds ahead instead of one second ahead and one second ahead.

  When the variable rate bit stream data predictively encoded by the encoding units 2′-1 and 2′-2 shown in FIG. 2 is transmitted via the network, the encoding side uses the transmission unit as shown in FIG. (Step S41), then a packet header is added (step S42), and then the packet is sent out on the network (step S43).

  As shown in FIG. 11A, the decoding side removes the header (step S51), then restores the data (step S52), then stores this data in the memory and waits for decoding (step S53). When decoding is performed, as shown in FIG. 11B, deformatting is performed (step S61), input / output control of the input buffer 22a is performed (step S62), and then unpacking is performed (step S61). S63). At this time, if there is a search reproduction instruction, the search pointer is decoded. Next, the predictor is selected and decoded based on the flag (step S64), then the input / output control of the output buffer 110 is performed (step S65), and then the original multi-channel is restored (step S66). This is repeated (step S67).

In the above embodiment, 2ch “1” and “2” related to the front group are set to “1” = Lf + Rf.
“2” = Lf−Rf
However, instead, the multi-channel is downmixed according to Equation (2) to generate stereo 2ch data (L, R),
Next, the following formula (1) ′
“1” = L + R
“2” = LR
“3” to “5” are the same. “6” = Lfe-C (1) ′
(2nd embodiment). In this case, the decoding-side mix & matrix circuit 4 ′ can generate channel R by adding channels “1” and “2” and subtracting channel L by subtraction.

  Also, as shown in FIG. 12 as the third embodiment, stereo 2ch data (L, R) is generated by down-mixing the multi-channel according to equation (2) instead of 2ch “1” and “2”. The stereo 2ch (L, R) and 4ch "3" to "6" may be predictively encoded. In the second and third embodiments, since the front left (Lf) and the front right (Rf) are not transmitted to the decoding side, they are generated by the equations (1) and (2) on the decoding side.

  Next, a fourth embodiment will be described with reference to FIGS. In the above embodiment, a group of correlated signals “1” to “6” is configured to be predictively encoded. In the fourth embodiment, a plurality of groups of correlated signals are generated. Thus, the prediction coding is performed, and the prediction coding data of the group having the highest compression rate is selected. For this reason, the encoding unit shown in FIG. 13 is provided with first to n-th correlation circuits 1-1 to 1-n, and these n correlation circuits 1-1 to 1-n are, for example, 6ch (Lf, C , Rf, Ls, Rs, Lfe) is converted into n types of 6-channel signals “1” to “6” having different correlations.

For example, the first correlation circuit 1-1 converts as follows:
“1” = Lf
“2” = C− (Ls + Rs) / 2
“3” = Rf−Lf
“4” = Ls−a × Lfe
“5” = Rs−b × Rf
“6” = Lfe
The n-th correlation circuit 1-n converts as follows.

“1” = Lf + Rf
“2” = C−Lf
“3” = Rf−Lf
“4” = Ls−Lf
“5” = Rs−Lf
“6” = Lfe-C
Further, a prediction circuit 15 and a buffer / selector 16 are provided for each of the correlation circuits 1-1 to 1-n, and the group having the highest compression rate is selected based on the data amount of the minimum value of the prediction residual for each group. It is selected by the signal generator 17b. At this time, the formatting circuit 19 adds and multiplexes the selection flag (correlation circuit selection flag, correlation coefficients a and b of the correlation circuit).

  Further, on the decoding side shown in FIG. 14, n correlation circuits 4-1 to 4-n (or coefficients a and b can be changed with respect to the correlation circuits 1-1 to 1-n on the encoding side. One correlation circuit (omitted) is provided. When the n groups of prediction circuits shown in FIG. 13 have the same configuration, the decoding device does not need to have n groups of prediction circuits as shown in FIG. Then, one of the correlation circuits 4-1 to 4-n is selected based on the selection flag transmitted from the encoding device, or the coefficients a and b are set and the original 6ch (Lf, C, Rf, Ls, Rs, Lfe) is restored, and the multi-channel is downmixed according to Equation (2) to generate stereo 2ch data (L, R).

  In the first embodiment described above, one type of correlation signal “1” to “6” is configured to be predictively encoded. The group of signals “1” to “6” A group of original signals (Lf, C, Rf, Ls, Rs, Lfe) may be predictively encoded, and a group with a higher compression rate may be selected.

It is a block diagram which shows 1st Embodiment of the audio | voice coding apparatus with which this invention is applied, and the audio | voice decoding apparatus corresponding to it. It is a block diagram which shows the encoding part of FIG. 1 in detail. It is explanatory drawing which shows the bit stream encoded by the encoding part of FIG. 1, FIG. It is explanatory drawing which shows the format of the pack of DVD. It is explanatory drawing which shows the format of the audio pack of DVD. It is a block diagram which shows the decoding part of FIG. 1 in detail. 7 is a timing chart showing write / read timings of the input buffer of FIG. 6. It is explanatory drawing which shows the compressed data amount for every access unit. It is explanatory drawing which shows an access unit and a presentation unit. It is a flowchart which shows the audio | voice transmission method. It is a flowchart which shows the audio | voice transmission method. It is a block diagram which shows 3rd Embodiment of the audio | voice encoding apparatus with which this invention is applied, and the audio | voice decoding apparatus corresponding to it. It is a block diagram which shows the audio | voice coding apparatus of 4th Embodiment. It is a block diagram which shows the audio | voice decoding apparatus of 4th Embodiment.

Explanation of symbols

1 '6ch mix & matrix circuit 13D1, 13D2, 15D1-15D4 Prediction circuit (buffer / selector 14)
The compression means is configured together with D1, 14D2, 16D1 to 16D4. )
14D1, 14D2, 16D1 to 16D4 Buffer / selector 17 Selection signal / DTS generator (timing generation means)
17c PTS generator (timing generator)
19 Formatting circuit (formatting means)
21 Deformatting circuit (separation means)
22 Unpacking circuit 22a Input buffer 24D1, 24D2, 23D1 to 23D4 Prediction circuit (expanding means)
100 Control unit (reading means)
110 Output buffer

Claims (2)

In response to a voice signal inputted audio signals of a plurality of channels each channel taken intact channel or correlated, with obtaining the top sample values by a plurality of linear prediction method different characteristics from the past in the time domain Selecting and predicting a linear prediction method such that a linear prediction value of a current signal is predicted, and a prediction residual obtained from the predicted linear prediction value and the speech signal is minimized;
Packing the prediction residual with the number of bits based on the number-of-bits information when packing the prediction encoded data including the linear prediction method, the prediction residual, and a predetermined head sample value of each selected channel ; When,
Generating decoding time stamp information indicating timing for reading compressed data in the input buffer on the decoding side according to the amount of the compressed data packed;
User data including a packet header including the decoding time stamp information, a compressed PCM private header including the UPC / EAN number and ISRC data of the voice signal, and a compressed PCM access unit including the compressed data. Formatting into packets;
A speech encoding method comprising: An audio decoding method for decoding the original audio signals of the plurality of channels from the data encoded by the audio encoding method according to claim 1,
Decoding the UPC / EAN number and ISRC data located in the compressed PCM private header;
Decoding the stored subpackets based on the decoding time stamp information to separate compressed PCM access units;
The prediction residual of the compressed data in the separated compressed PCM access unit is decoded with the number of bits based on bit information, and the prediction value is calculated based on the decoded prediction residual, the head sample value, and the linear prediction method. Calculating for each channel ;
Restoring the original audio signals of the plurality of channels from the calculated predicted values;
A speech decoding method comprising:

Images