JP3346384B2 - Audio coding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a speech coding method for compressing a multi-channel speech signal.

[0002]

2. Description of the Related Art As a method for compressing an audio signal, the present inventor has disclosed in Japanese Patent Application No.
A plurality of predictors having different characteristics are used to calculate a plurality of linear prediction values of a current signal from a past signal in a time domain for an original digital audio signal of a channel, and the original digital audio signal and the plurality of linear prediction values are calculated. A prediction coding method for calculating a prediction residual for each predictor from, and selecting a minimum value of the prediction residual is proposed.

In the above method, the original digital audio signal has a sampling frequency = 96 kHz and the number of quantization bits = 2.
Although a certain degree of compression effect can be obtained in the case of about 0 bits, in recent DVD audio discs, this 2
A double sampling frequency (= 192 kHz) is used, and the number of quantization bits tends to be 24 bits.

[0004]

As described above, in the case of DVD audio, there are a plurality of sampling frequencies of 48 kHz, 96 kHz and 192 kHz. And encoded,
It is desirable to be able to perform a simplified decoding process. Therefore,
It is an object of the present invention to provide a speech encoding method that takes such points into consideration.

[0005]

The present invention, in order to achieve the above object, comprises the following means. That is,

The number of samples is determined in accordance with a predetermined sampling frequency of an incoming audio signal, and a multi-channel audio signal quantized with the determined number of samples is used for each channel as it is or for each channel correlated with each other. In addition, the first sample value is obtained in response to the input audio signal, and the linear prediction values of the current signal from the past in the time domain are respectively predicted by a plurality of linear prediction methods having different characteristics.
From the predicted linear prediction value and the audio signal
A step of selecting a linear prediction method that minimizes the obtained prediction residual and performing predictive encoding; and a step of selecting a linear prediction method and a prediction residual for each channel selected by the step.
A subpacket storing predicted encoded data including the first sample value of the original analog audio signal on the reproduction side.
Step a speech encoding method consisting of formatting the data structure having a synchronization information part including the sampling frequency and the quantization bit used in is restored.

[0007]

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 speech encoding apparatus to which the present invention is applied and a speech decoding apparatus corresponding to the speech encoding apparatus. FIG. 2 is a block diagram showing the encoding unit of FIG. 1 in detail. FIG. 1 is an explanatory diagram showing a bit stream encoded by the encoding unit shown in FIG. 2, and FIG.
FIG. 5 is an explanatory diagram showing the format of a DVD audio pack, FIG. 6 is an explanatory diagram showing the format of the audio data area of FIG. 5 in detail, and FIG. 7 is a diagram showing the decoding unit of FIG. FIG. 8 is a detailed block diagram, FIG. 8 is a timing chart showing write / read timing of the input buffer of FIG. 7, FIG. 9 is an explanatory diagram showing the amount of compressed data for each access unit, and FIG. It is.

Here, for example, the following four systems are known as multi-channel systems. (1) Four-channel system As in the Dolby surround system, a total of four channels including three channels of front L, C, and R + one channel of rear S (2) Five-channel system S in the Dolby AC-3 system
Like without W channel, 3 channels of front L, C and R + 2 channels of rear SL and SR, total 5 channels (3) 6 channel system DTS (Digital Theater)
6) (L, C, R, SW (Lfe), SL, SR) like the Dolby AC-3 system (4) 8-channel system SDDS (Sony Dynamic D
digital sound), forward L, LC, C, R
6 channels of C, R, SW + 2 channels of rear SL, SR, total 8 channels

The 6-channel (ch) mix & matrix circuit 1 'on the encoding side shown in FIG. 1 includes a front left (Lf), a center (C), a front right (Rf), a surround left ( Ls), surround light (Rs) and Lfe (Low Frequency Effe)
ct), the 6-channel PCM data is classified and converted into 2ch “1” and “2” for the front group and 4ch “3” to “6” for the other group by the following equation (1), and converted into 2ch “1”.
"2" is assigned to the first encoding unit 2'-1, and 4ch "3" to
"6" is 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 where 0 ≦ a ≦ 1 (1)

As shown in detail in FIG. 2, the first and second encoders 2'-1 and 2'-2 which constitute the encoder 2 'respectively have 2ch "1", "2" and 4ch "3". "~" 6 "PC
The M data is predictively encoded for each channel, and the encoded prediction data is transmitted as a bit stream as shown in FIG. 3 to the decoding side via a recording medium 5 or a communication medium 6 such as a satellite line or a telephone line. On the decoding side, the first and second decoding units 3'-1 and 3'-2 constituting the decoding unit 3 'respectively perform 2ch "1" for the forward group, as shown in detail in FIG.
The prediction coded data of "2" and 4ch "3" to "6" relating to the other groups are decoded into PCM data for each channel.

Then, the original 6 ch (Lf, C, Rf, Ls,
Rs, Lfe) are restored, and stereo 2-ch data (L, R) is generated from the original 6 ch and coefficients mij (i = 1, 2, j = 1, 2 to 6) as in the following equation (2). I do. 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)

Referring to FIG. 2, encoding sections 2'-1, 2'-
2 will be described in detail. PC for each channel "1" to "6"
The M data is stored in one frame buffer 10 for each frame. Then, the sample data of each of the channels “1” to “6” of one frame are respectively supplied to the prediction circuits 13D1 and 13D
2, 15D1 to 15D4 and each channel
First sample data (stored in a restart header described later) of each frame of “1” to “6” is applied to the unpacking circuit 8 and the formatting circuit 19. The sampling frequency (fs) and the number of quantization bits (Qb) when the PCM data is A / D converted are applied to the packing circuit 18 and the formatting circuit 19.
The prediction circuits 13D1, 13D2, and 15D1 to 15D4 respectively calculate the PCM data of each channel “1” to “6”.
A plurality of predictors (not shown) having different characteristics calculate a plurality of linear prediction values of the current signal from a past signal in the time domain, and then perform prediction for each predictor from the original PCM data and the plurality of linear prediction values. Calculate the residual. The following buffer
The selectors 14D1, 14D2, 16D1 to 16D4 are prediction circuits 13D1, 13D2, 15D1 to 15D, respectively.
4 is temporarily stored, and the minimum value of the prediction residual is selected for each subframe specified by the selection signal / DTS (decoding time stamp) generator 17.

The selection signal / DTS generator 17 applies the bit number flag of the prediction residual to the packing circuit 18 and the formatting circuit 19, and outputs a predictor selection flag indicating the predictor having the minimum prediction residual. To the formatting circuit 19, the correlation coefficient a in the equation (1) and the DTS indicating the time at which the decoding side extracts the stream data from the input buffer 22a (FIG. 7). Packing circuit 18
Are buffer / selectors 14D1, 14D2, 16D1-1.
The prediction residual for 6 ch selected by 6D4 is packed with the specified number of bits based on the bit number flag specified by the selection signal / DTS generator 17, and packed with the number of samples corresponding to the sampling frequency fs. That is, when the sampling frequency is 48 kHz, 40 samples are selected, when the sampling frequency is 96 kHz, 80 samples are selected, and when the sampling frequency is 192 kHz, 80 samples are selected. Further, the PTS generator 17c generates a PTS (presentation time stamp) indicating a time at which the decoding side takes out the PCM data from the output buffer 110 (FIG. 7) and outputs the PTS to the formatting circuit 19.

The following formatting circuit 19 is shown in FIGS.
Format as user data as shown in FIG.
Is a variable-rate bit stream (sub-stream) BS0 including 2ch “1” and “2” prediction coded data related to the forward group.
And a variable-rate bit stream (substream) BS1 including 4ch “3” to “6” prediction coded data relating to other groups, and a bitstream header (restart header) provided before substreams BS0 and BS1 ). Also, substream B
One frame of S0 and BS1 includes: a frame header; first sample data of one frame of each channel “1” to “6”; and selection of a predictor for each subframe of each channel “1” to “6”. A flag; a bit number flag for each subframe of each ch “1” to “6”; a prediction residual data string (variable bit number) for each ch “1” to “6”; Are multiplexed. According to such predictive coding, the original signal has, for example, a sampling frequency (fs) = 96 kHz and a quantization bit number (Qb) =
In the case of 24 bits and 6 channels, a compression ratio of 71% can be realized.

When the variable rate bit stream data predictively encoded by the encoding units 2'-1 and 2'-2 shown in FIG. 2 is recorded on a DVD audio disc as an example of a recording medium, FIG. The audio (A) pack shown is packed. This pack has 4 bytes of pack start information and 6 bytes of SCR (Syst) for 2034 bytes of user data (A packet, V packet).
em Clock Reference (system time reference value) information, a 3-byte Mux rate (rate) information, and a 1-byte stuffing that add a pack header of a total of 14 bytes (1 pack = 2048 bytes in total) . In this case, the time stamp SCR information is
In the first pack, the time of the A-pack in the same title can be managed by setting it to “1” so as to be continuous within the same title.

As shown in detail in FIG. 5, the A packet of the compressed PCM has a packet header of 9 to 22 bytes and a compressed P packet.
The CM private header and 1 to 2015 bytes of audio data (compressed P
CM). And DTS and PTS
5 is set in the packet header of FIG. 5 (specifically, the PTS is set in the 10th to 14th bytes and the DTS is set in the 15th to 19th bytes). The private header of the compressed PCM is: 1-byte substream ID, 2 bytes of UPC / EAN-ISRC (Universal Prism).
oduct Code / European Article Number-International S
tandard Recording Code) number and UPC / EAN-
ISRC data, 1-byte private header length, 2-byte first access unit pointer, 4-byte audio data information (ADI), and 0-7 byte stuffing bytes. ing.

The forward access unit search pointer for searching for an access unit one second later and the backward access unit search pointer for searching for an access unit one second earlier in the ADI are both 1
Set in bytes. Specifically, the forward access unit search pointer is set in the first byte of the ADI, and the backward access unit search pointer is set in the eighth byte. Thus, ADI can store up to 2015 bytes of audio data in order to reduce it to 4 bytes in compressed PCM.

The audio data area in the compressed PCM (PPCM) audio packet shown in FIG. 5 is composed of a plurality of PPCM access units as shown in FIG. 6, and the PPCM access unit is composed of PPCM sync information and sub-packets. I have. First PP
The subpacket in the CM access unit includes a directory, a substream “BS0”, a CRC (1 byte or 2 bytes), a substream “BS1”,
C and extra information, and the sub-streams “BS0” and “BS1” are composed of only PPCM blocks. Sub-packets in the second and subsequent PPCM access units also include a directory, a sub-stream “BS0”, a CRC, and a sub-stream “BS1”.
, CRC and extra information, and the sub-streams “BS0” and “BS1” have a restart header and P
It is composed of PCM blocks. The extra information has at least a size adjustment function. That is, when the incoming data has a fixed rate (CBR), the number of samples per packet is 40, 80, 1 depending on the sampling frequency fs as described above.
60, the data length per packet may not match the size of the subpacket depending on the determined number of samplings. In order to match it with the size of the subpacket, for example, ,
The size is adjusted by adding 0, 0, etc. In addition, text data or the like can be used as the data for adjusting the size.

The PPCM sync information (hereinafter also referred to as synchronization information) includes the following information. -Number of samples per packet: 40, 80 or 160 is selected according to the sampling frequency fs. If the data rate is VBR, "0" (an identifier indicating that the data in the subpacket is compressed data of VBR); if the data rate is CBR, "1" (the data in the subpacket has a fixed rate) Identifier indicating that sampling frequency fs and number of quantization bits Qb Channel assignment information

Next, referring to FIG. 7, the decoding units 3'-1,
3′-2 will be described. The variable rate bit stream data BS0 and BS1 in the above format are separated by the deformatting circuit 21. And each channel
The head sample data of one frame of "1" to "6" and the predictor selection flag are respectively stored in the prediction circuits 24D1 and 24D.
2, 23D1 to 23D4, and each channel “1” to
The bit number flag of “6” is applied to the unpacking circuit 22. The SCR, the DTS, and the prediction residual data string are applied to the input buffer 22a, and the PTS is applied to the output buffer 110. The channel allocation information is transmitted to the unpacking circuit 22 through the input buffer 22a.
And the prediction circuits 24D1, 24D2, 23D1 to 23 used in accordance with the channel assignment information.
D4 is determined. For example, in the case of allocation information of 6 channels, all prediction circuits are used, and data of 6 channels is repeatedly supplied. In the case of allocation information of 5 channels, data of 5 channels is repeatedly supplied. I have. The allocation information is not limited to multi-channel information, but may be stereo 2-channel information. In this case, the prediction circuits 24D1 and 24D2 are used. An identifier indicating whether the data rate is VBR or CBR is stored in each predictor 24.
D1, 24D2, 23D1, 23D2, 23D3, 23
D4, the processing program for input / output data corresponding to the identifier is determined and processed. In the case of VBR, it is necessary to switch the processing program and to load the input data every time, so that it takes time for processing. In the case of CBR, the processing rate does not need to be switched because the processing rate is fixed because the processing rate is fixed. Be faster. The sampling frequency fs and the number of quantization bits Qb are applied to the D / A converter 102. Here, the prediction circuits 24D1, 24D2, 23D1
A plurality of predictors (not shown) in the 23D4 are prediction circuits 13D1, 13D2, 15D1 to 15D on the encoding side, respectively.
4 have the same characteristics as those of the plurality of predictors, and those having the same characteristics are selected by the predictor selection flag.

An audio packet is first separated from the audio pack by the reformatting circuit 21. Next, stream data (predicted residual data string) is separated from the audio packet, and the bit streams BS0 and BS0 are separated.
1 is taken out. Further, the SCR is taken out and taken into the input buffer 22a for each access unit according to the timing by the SCR as shown in FIG. Here, the data amount of one access unit is
For example, when fs = 96 kHz (1/96 kHz)
The second is a variable length as shown in FIGS. 9 and 10A in detail. Then, 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 is provided for each channel "1" to
The prediction residual data string of “6” is separated based on each bit number flag, and is divided into prediction circuits 24D1, 24D2, and 23, respectively.
It outputs to D1-23D4. Prediction circuits 24D1, 24D
2, 23D1 to 23D4, the current prediction residual data of each of the channels “1” to “6” from the unpacking circuit 22 and each of the plurality of internal predictors selected by the predictor selection flag. The previous predicted value predicted by one frame is added to calculate the current predicted value, and then the PC of each sample is determined based on the first sample data of one frame.
M data is calculated and stored in the output buffer 110. PCM data stored in the output buffer 110 is P
The data is read out and output based on the TS. Therefore, the variable-length access unit shown in FIG. 10A is expanded, and a fixed-length presentation unit shown in FIG. 10B is output.

Further, based on the sampling frequency fs and the number of quantization bits Qb in the PPCM sync information, the PC
The M data is converted by the D / A converter 102 into an analog signal. At the same time, C
When the BR identifier is detected, the position of the extra data in the directory is detected, and further, for example, data of 0, 0... Or extra data for size adjustment such as text data is detected, it is converted to text data. In some cases, the extra data is supplied from the unpacking circuit 22 to a text data decoding circuit (not shown), where the data is decoded, extracted as text data, and output through the output buffer 110. On the other hand, if the extra data is 0,0... Data, no processing is performed. If the text data decoder circuit is not prepared, this processing is passed. Also, where
When a search reproduction is instructed via the operation unit 101, the control unit 100 performs a search based on the forward access unit search pointer (one second ahead) and the backward access unit search pointer (one second before) shown in FIG. Play the access unit. As this search pointer, one second ahead,
Instead of one second before, two seconds before and two seconds before may be used.

When variable-rate bit stream data predictively coded by the coding units 2'-1 and 2'-2 shown in FIG. 2 is transmitted through a network, the coding side performs the processing shown in FIG. (Step S41), add a packet header (step S42), and send this packet out onto the network (step S43).

On the decoding side, as shown in FIG. 12A, the header is removed (step S51), the data is restored (step S52), and the data is stored in the memory to wait for decoding (step S53). . Then, when decoding is performed, as shown in FIG. 12B, deformatting is performed (step S61), and then the input buffer 22
The input / output control of a is performed (step S62), and then the unpacking is performed (step S63). At this time, if there is a search reproduction instruction, the search pointer is decoded. Next, a predictor is selected and decoded based on the flag (step S64), input / output control of the output buffer 110 is performed (step S65), and the original multi-channel is restored (step S66). This is output (step S67), and thereafter, this is repeated.

In the above embodiment, 2ch "1" and "2" related to the front group are transformed by "1" = Lf + Rf "2" = Lf-Rf and are predictively coded. , Down-mixing the multi-channels to generate stereo 2-ch data (L, R), and then the following equation (1) ′ “1” = L + R “2” = LR “3” to “5” are the same as “6”. = Lfe-C (1) ′ and may be subjected to predictive coding (second
Embodiment). In this case, the mix & matrix circuit 4 ′ on the decoding side adds the channels “1” and “2” to the channel L, and subtracts the channel R by adding the channels “1” and “2”.
Can be generated.

As a third embodiment, as shown in FIG. 13, instead of 2ch "1" and "2", multi-channels are downmixed by equation (2) and stereo 2ch data (L, R) is obtained. The stereo 2ch (L, R) and the 4ch “3” to “6” may be generated and predictively coded. In the second and third embodiments, since the front left (Lf) and the front right (Rf) are not transmitted to the decoding side, the decoding side generates them according to equations (1) and (2).

Next, a fourth embodiment will be described with reference to FIG. 14, FIG. 15, and FIG. In the above embodiment,
Although one group of correlated signals “1” to “6” are configured to be predictively coded, in the fourth embodiment, a plurality of groups of correlated signals are generated and predictively coded, It is configured to select the prediction coded data of the group having the highest compression ratio. In this embodiment, the encoding in one group is not performed by classifying and converting into 2ch for the front group and 4ch for the other group as in the above-described embodiments. FIG. 1 shows a configuration in which the combined encoding process is performed.
4 is shown as a diagram corresponding to FIG. Also,
FIG. 15 shows a detailed block of the encoding unit.
In this embodiment, n correlation circuits 1-1 to 1-n are provided on the mix & matrix circuit 1 'side.
These n correlation circuits 1-1 to 1-n are, for example, 6 ch (L
f, C, Rf, Ls, Rs, Lfe) are 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 Further, the n-th correlation circuit 1-n performs conversion as follows. “1” = Lf + Rf “2” = C−Lf “3” = Rf−Lf “4” = Ls−Lf “5” = Rs−Lf “6” = Lfe−C

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 ratio is determined based on the data amount of the minimum prediction residual for each group. Are selected by the correlation selection 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).

FIG. 16 shows a data area corresponding to FIG. 6 described above. In this embodiment, the sub-stream "B"
Instead of using “S1”, the sub-stream “BS0” alone is used.

On the decoding side shown in FIG. 17, n correlator circuits 4 are provided for the correlator circuits 1-1 to 1-n on the encoding side.
−1 to 4-n (or one correlation circuit (not shown) whose coefficients a and b can be changed) are provided. Note that n shown in FIG.
When the prediction circuits of the groups have the same configuration, the decoding device does not need to provide the prediction circuits of n groups as shown in FIG. 17, and may use the prediction circuits of one group. 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 selected.
And set the original 6 ch (Lf, C, Rf, Ls, Rs, L
fe), and down-mixes the multi-channels according to equation (2) to generate stereo 2-ch data (L, R).

In the first embodiment, one kind of correlation signal "1" to "6" is configured to be predictively coded, but the signals "1" to "6" are encoded. And the group of the original signals (Lf, C, Rf, Ls, Rs, Lfe) may be predictively coded and the group with the higher compression ratio may be selected.

[0035]

As described above, according to the present invention, the number of samples is determined in accordance with the sampling frequency so that the length of the audio signal during reproduction is constant. Also in the case of processing the audio signal of the sampling frequency, the processing on the reproduction side can be simplified.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of a speech encoding device to which the present invention is applied and a speech decoding device corresponding thereto.

FIG. 2 is a block diagram illustrating an encoding unit of FIG. 1 in detail.

FIG. 3 is an explanatory diagram showing a bit stream encoded by an encoding unit shown in FIGS. 1 and 2;

FIG. 4 is an explanatory diagram showing a format of a DVD pack.

FIG. 5 is an explanatory diagram showing a format of a DVD audio pack.

6 is an explanatory diagram showing the format of the audio data area in FIG. 5 in detail.

FIG. 7 is a block diagram illustrating a decoding unit of FIG. 1 in detail.

FIG. 8 is a timing chart showing write / read timings of the input buffer of FIG. 7;

FIG. 9 is an explanatory diagram showing the amount of compressed data for each access unit.

FIG. 10 is an explanatory diagram showing an access unit and a presentation unit.

FIG. 11 is a flowchart showing a voice transmission method.

FIG. 12 is a flowchart showing a voice transmission method.

FIG. 13 is a block diagram showing a third embodiment of a speech encoding device to which the present invention is applied and a speech decoding device corresponding thereto.

FIG. 14 is a block diagram showing a third embodiment of a speech encoding device to which the present invention is applied and a speech decoding device corresponding thereto.

FIG. 15 is a block diagram illustrating a speech encoding device according to a fourth embodiment.

FIG. 16 is an explanatory diagram of another embodiment corresponding to FIG.

FIG. 17 is a block diagram illustrating a speech decoding device according to a fourth embodiment.

[Explanation of symbols]

1 '6ch Mix & Matrix Circuit 13D1, 13D2, 15D1-15D4 Prediction Circuit (Buffer / Selector 14D1, 14D2, 16D1-1
A compression means is constituted together with 6D4. 14D1, 14D2, 16D1-16D4 buffer
Selector 17 Selection signal / DTS generator (timing generating means) 17c PTS generator (timing generating means) 19 Formatting circuit (Formatting means) 21 Deformatting circuit (Separating means) 22 Unpacking circuit 22a Input buffer 24D1, 24D2, 23D1 to 23D4 Prediction circuit (expansion means) 100 Control unit 102 D / A converter 110 Output buffer

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-64-44499 (JP, A) JP-A-2-127899 (JP, A) JP-A-3-24834 (JP, A) JP-A 8- 272393 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 19/00-19/14 G11B 20/10-20/12 H03M 7/30-7/40 H04S 3/00 -3/02

Claims (1)

(57) [Claims] The number of samples is determined in accordance with a predetermined sampling frequency of an incoming audio signal, and the multi-channel audio signal quantized with the determined number of samples is correlated with the channel as it is or with each other. For each channel, the first sample value is obtained in response to the input audio signal, and the linear prediction values of the current signal from the past in the time domain are predicted by a plurality of linear prediction methods having different characteristics.
From the predicted linear prediction value and the audio signal
A step of selecting a linear prediction method that minimizes the obtained prediction residual and performing predictive encoding; and a method of linear prediction for each channel selected in the step.
A sub packet storing the predictive coding data including the Law and prediction residual and a predetermined first sample value of the original at the reproducing side
Step a speech encoding method consisting of formatting the data structure having a synchronization information part including a sampling <br/> ring frequency and quantization bit used when being restored to the analog audio signal.