JP2001188569A - Voice coding method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve compression ratio in the case or predictive-coding of voice signals. SOLUTION: An adding circuit 1a calculates a sum signal (L+R) of two- channel stereo signals L, R, and a subtracting circuit 1b calculates a difference signals (L-R). Difference arithmetic circuits 11D1, 11D2 calculate difference Δ(L+R), Δ(L-R) between this time and least time, and prediction-coding circuits (15D1, 15D2, 16D1, 16D2) calculate plural predicted values of the differences Δ(L+R), Δ(L-R), and calculate each prediction residual of the plural predicted values and the differences Δ(L+R), Δ(L-R), and a minimum prediction residual is selected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech coding method for predictively coding and compressing a speech signal.

[0002]

2. Description of the Related Art As a method of predictive encoding of a speech signal,
In the prior application (Japanese Patent Application No. 9-289159), the present inventor has applied a plurality of predictors having different characteristics to an original digital audio signal of one channel (channel) from a past signal in the time domain to a current signal. A method of calculating a plurality of linear prediction values, calculating a prediction residual for each predictor from the original digital audio signal and the plurality of linear prediction values, and selecting a minimum value of the plurality of prediction residuals; I have.

[0003]

However, in the above method, the original digital audio signal has a sampling frequency = 96.
In the case of kHz and the number of quantization bits = about 20 bits, a certain compression effect can be obtained.
In audio discs, twice this sampling frequency (= 192 kHz) is used, and the number of quantization bits tends to be 24 bits. Therefore, it is necessary to improve the compression ratio.

Accordingly, an object of the present invention is to provide a speech coding method capable of improving the compression ratio when predictive coding of a speech signal is performed.

[0005]

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

[0006] The first and second two of the same sampling frequency
Mixing the audio signals of the systems and converting them into channels correlated with each other; obtaining the first sample value in response to the input audio signal for each channel by converting the two correlated signals converted in the step; A step of selecting a linear prediction method whose prediction residual is a minimum value among a plurality of prediction values of a current signal predicted from a past signal in the time domain and performing predictive encoding, header information and user data; In the sub-packet in the compressed PCM access unit in the compressed PCM data format having the compressed PCM access unit in the user data, the first sample value selected by the step, the prediction residual, and the linear prediction method are stored. And storing the predicted data including the data.

[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 speech encoding method according to the present invention is applied and a speech decoding apparatus corresponding thereto, and FIG. 2 is a block diagram showing the encoder of FIG. 1 in detail. 3 is an explanatory diagram showing the format of one frame multiplexed by the multiplexer of FIG. 2, FIG. 4 is an explanatory diagram showing the format of a DVD pack, FIG. 5 is an explanatory diagram showing the format of a DVD audio pack, and FIG. FIG. 2 is a block diagram illustrating a decoder of FIG. 1 in detail.

The channel correlation circuit A shown in FIG. 1 has an addition circuit 1a and a subtraction circuit 1b. In the addition circuit 1a, each channel (hereinafter, ch) has a sampling frequency of 192 k
Calculates the sum signal (L + R) of the stereo 2ch signals L and R with the frequency and the number of quantization bits = 24 bits and outputs the sum signal to the 1ch lossless encoder 2D1 for the sum channel, and the subtraction circuit 1b calculates the difference signal (LR). The calculated value is output to the 1ch lossless encoder 2D2 for the difference channel. As shown in detail in FIG. 2, the encoders 2D1 and 2D2 predictively encode differences Δ (L + R) and Δ (LR) of the sum signal (L + R) and the difference signal (LR), respectively, and Transmitted via.

On the decoding side, the decoders 3D1 and 3D2 decode the prediction coded data of each channel into a sum signal (L + R) and a difference signal (LR), respectively, as shown in detail in FIG. Circuit B outputs the sum signal (L +
R) and the difference signal (LR) are restored to stereo 2-ch signals L and R.

Referring to FIG. 2, encoders 2D1, 2D2
Will be described in detail. The sum signal (L + R) and the difference signal (LR) are stored in one frame buffer 10 for each frame. Then, each sample value (L +
R) and (LR) are the difference calculation circuits 11D1, 1D1,
1D2, the difference Δ (L + R), Δ
(LR), that is, differential PCM (DPCM) data is calculated. Also, the first sample value (L +
R) and (LR) are applied to the multiplexer 19.

The difference Δ (L + R) calculated by the difference calculation circuit 11D1 is calculated by a plurality of predictors 12 having different prediction coefficients.
a-1 to 12a-n and subtracters 13a-1 to 13a-n. Then, the predictors 12a-1 to 12a-n calculate the respective predicted values of the difference Δ (L + R) based on the respective prediction coefficients, and the subtractors 13a-1 to 13b-n calculate the respective predicted values and the difference Δ Each prediction residual of (L + R) is calculated. The buffer / selector 16D1 temporarily stores the plurality of prediction residuals, selects the minimum prediction residual for each subframe specified by the selection signal generator 17, and outputs the selected prediction residual to the packing circuit 18. Note that this subframe has a sample length of about one-tenth of a frame, and one frame is, for example, 80 subframes. here,
Predictors 12a-1 to 12a-n and subtracters 13a-1 to 13a-1
3a-n constitute a prediction circuit 15D1 for the sum signal ch, and the prediction circuit 15D1 and the buffer / selector 16D1
Constitutes a predictive encoding circuit for the sum signal ch.

Similarly, the difference Δ (LR) calculated by the difference calculation circuit 11D2 is calculated by using a plurality of predictors 12b-1 to 12b-n and subtracters 13b-1 to 13b-13 having different prediction coefficients.
b-n. Then, the predictors 12b-1 to 12b-12
bn, the difference Δ (L−
R) are calculated, and the subtractors 13b-1 to 13b
In −n, each prediction residual of the difference Δ (LR) from each of the prediction values is calculated. The buffer / selector 16D2 temporarily stores the plurality of prediction residuals, and
The minimum prediction residual is selected for each sub-frame specified by, and is output to the packing circuit 18. Predictor 12b-
1 to 12b-n and the subtractors 13b-1 to 13b-n form a prediction circuit 15D2 for the difference signal ch, and the prediction circuit 15D2 and the buffer / selector 16D2 form a prediction encoding circuit for the difference signal ch. are doing.

The selection signal generator 17 applies a bit number flag (5 bits) of the prediction residual to the packing circuit 18 and the multiplexer 19, and a predictor selection flag (predictor) indicating the predictor having the minimum prediction residual. The number n is 2 to 9 and 3 bits) are applied to the multiplexer 19. The packing circuit 18 includes buffer / selectors 16D1, 16D
The prediction residual for 2 ch selected by 2 is packed with the specified number of bits based on the bit number flag specified by the selection signal generator 17.

As shown in FIG. 3, the following multiplexer 19 outputs a frame header (40 bits), a first sample value (25 bits) of one frame of the sum signal ch (L + R) for one frame, A head sample value (25 bits) of one frame of the signal ch (LR); a predictor selection flag (3 bits × 80) for each subframe of the sum signal ch (L + R); a difference signal ch (L) −R) a predictor selection flag (3 bits × 80) for each subframe; a sum signal ch (L + R) a bit number flag for each subframe (5 bits × 80); a difference signal ch (L− R) the number-of-bits flag (5 bits × 80) for each sub-frame; a prediction residual data sequence (variable number of bits) of the sum signal ch (L + R); and a prediction residual of the difference signal ch (LR). Difference data string (variable bit ) And multiplexed as an access unit, output as a variable rate bit stream. The prediction residual data sequence forms a subpacket. According to such predictive coding, the original signal is, for example, sampling frequency = 192.
In the case of kHz, the number of quantization bits = 24 bits, and two channels, a compression ratio of 59% can be realized.

When recording the variable rate bit stream data on a DVD audio disk,
It is packed in the audio (A) pack of the compressed PCM shown in FIG. This pack has 4 bytes of pack start information for 2034 bytes of user data (A packet and V packet) and 6 bytes of SCR (System Clock).
Reference: system time reference value, 3-byte Mux rate (rate) information, and 1-byte stuffing for a total of 14-byte pack header (1 pack = 2048 bytes in total). In this case, the time of the A pack in the same title can be managed by setting the SCR information as the time stamp to be “1” in the first pack in the ACB unit so as to be continuous in the same title.

As shown in detail in FIG. 5, the A packet of the compressed PCM is composed of a packet header of 17, 9, or 14 bytes, a private header, and audio compressed PCM data of 1 to 2015 bytes in the format shown in FIG. ing. 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 bytes of audio data information (ADI), and 0 to 7 bytes of stuffing bytes. ing. Thus, the AD of the A packet of the compressed PCM
I is chosen to be 4 bytes and is the normal uncompressed PCM A
It is shorter by 4 bytes than the ADI of the packet.
Therefore, the audio data can be increased by 4 bytes.

Next, referring to FIG. 6, decoders 3D1, 3D
2 will be described. The variable rate bit stream data of the format shown in FIG.
1 based on the frame header. Then, the leading sample values of one frame of the sum signal ch (L + R) and the difference signal ch (LR) are respectively calculated by the accumulation operation circuit 2.
5a and 25b, the predictor selection flags of the sum signal ch (L + R) and the difference signal ch (LR) are set to predictors (24a-1 to 24a-n) and (24b-1 to 24b, respectively).
−n), and the sum signal ch (L +
R) and the bit number flag of the difference signal ch (LR) and the prediction residual data string are applied to the unpacking circuit 22.
Here, the predictors (24a-1 to 24a-n), (24b
-1 to 24b-n) are predictors (1) on the encoding side, respectively.
2a-1 to 12a-n), (12b-1 to 12b-n)
And the same characteristic is selected by the predictor selection flag.

The unpacking circuit 22 outputs a sum signal ch (L
+ R) and the prediction residual data sequence of the difference signal ch (LR) are separated based on the bit number flags and output to the adders 23a and 23b, respectively. The adder circuits 23a and 23b respectively add the sum signal ch from the unpacking circuit 22.
(L + R) and the current prediction residual data of the difference signal ch (LR), and the predictors (24a-1 to 24a-n), (24
b-1 to 24b-n), the previous predicted value predicted by each one selected by the predictor selection flag is added to calculate the current predicted value. This forecast is
The differences Δ (L + R) and Δ (LR) calculated by the difference circuits 11a and 11b shown in FIG.
M data, predictors (24a-1 to 24a-n),
(24b-1 to 24b-n) and cumulative operation circuits 25a, 2
5b.

The cumulative operation circuits 25a and 25b are respectively
The difference Δ (L +
R) and Δ (LR) are cumulatively added for each sample to output PCM data of a sum signal ch (L + R) and a difference signal ch (LR). The sum signal (L + R) and the difference signal (L−
As for R), as shown in FIG. 1, a 2L signal is calculated by the adding circuit 4a, and a 2R signal is calculated by the subtracting circuit 4b. Then, the 2L signal and the 2R signal are each divided by 1/2 by the dividers 5a and 5b, and the original stereo two-channel signals L and R are restored.

Next, a second embodiment will be described with reference to FIGS. In the above embodiment, the sum signal (L +
R), each difference Δ (L + R), Δ (L) of the difference signal (LR)
-R), that is, predictive encoding is performed only on the DPCM data. In the second embodiment, the sum signal (L + R), the difference signal (LR), ie, the PCM data, or each difference Δ (L + R), Δ (LR), that is, DPCM data is selectively and predictively encoded.

For this reason, the encoding apparatus shown in FIG.
The sum signal (L + R) and the difference signal (L−
R) for predictive encoding, respectively.
15S and buffer / selectors 16A and 16S are added. The selection signal generator 17 is a buffer / selector 1
6A, the sum signal (L +
R), the difference signal (LR) and the buffer / selector 16D
1, ΔD (L +
R), Δ (LR) based on the minimum value of each prediction residual.
It is determined whether the compression ratio of PCM data or DPCM data is higher, and the higher data is selected. At this time, the PCM / DPCM selection flag (prediction circuit selection flag) is added and multiplexed.

Here, the prediction circuit 15A for the sum signal (L + R) and the prediction circuit 15D1 for the difference Δ (L + R) shown in FIG. 7 have the same configuration, and the prediction circuit 15S for the difference signal (L−R) has the same configuration. When the difference Δ (LR) prediction circuits 15D2 have the same configuration, the decoding apparatus uses the PCM as shown in FIG.
It is not necessary to provide a prediction circuit for both data and DPCM data, and a prediction circuit for one data may be used. Then, based on the prediction circuit selection flag transmitted from the encoding device, the selectors 26a and 26b select the outputs of the accumulator circuits 25a and 25b in the case of DPCM data.
In the case of M data, the outputs of the adders 23a and 23b are selected.

In the third embodiment, as shown in FIG. 9, original signals L and R (PCM data), a sum signal (L + R), a difference signal (LR) (PCM data), and each difference Δ (L
+ R) and one of the three groups Δ (LR) (DPCM data) are selectively predictively encoded.

For this reason, in the encoding device shown in FIG.
In addition to the configuration shown in (1), prediction circuits 15L and 15R for predictively encoding the original signals L and R, and buffers / selectors 16L and 16R are added. Further, the selection signal generator 17 includes the original signals L and R selected by the buffer / selectors 16L and 16R and the buffer / selectors 16A and 16S.
Signal (L + R), difference signal (LR) selected by
And data of a group having a high compression rate based on the minimum values of the prediction residuals of the differences Δ (L + R) and Δ (LR) selected by the buffer / selectors 16D1 and 16D2. At this time, the selection flag (prediction circuit selection flag) is added and multiplexed.

When the three groups of prediction circuits shown in FIG. 9 have the same configuration, the decoding device does not need to provide three groups of prediction circuits as shown in FIG. Is fine. Then, based on the prediction circuit selection flag transmitted from the encoding device, the output of the accumulation operation circuits 25a and 25b is selected in the case of DPCM data, and the addition circuits 23a and 23b in the case of PCM data.
And the channel correlation circuit B selects the original signal L,
Restore R. Further, in the case of the group of the original signals L and R by the selectors 27a and 27b, the addition circuit 23
a, 23b are selected, and in other cases, the output of the channel correlation circuit B is selected.

When the variable-rate bit stream data predicted and coded by the coding side is transmitted through a network, the coding side packetizes the data for transmission as shown in FIG. 11 (step S41). Next, a packet header is added (step S42), and then this packet is sent out onto the network (step S4).
3). On the decoding side, the header is removed as shown in FIG. 12 (step S51), the data is restored (step S52), and the data is stored in a memory and decoding is waited (step S53).

FIG. 13 shows an aspect of the compressed PCM (PPCM) audio (A) packet shown in FIG. 5 which is different from FIG. In this different aspect, the audio data area in the audio (A) packet of the compressed PCM (PPCM) is composed of a plurality of PPCM access units as shown in FIG.
It is composed of CM sync information and subpackets.
The subpacket in the first PPCM access unit is:
Directory, bitstream BS0, CRC,
Bit stream BS composed of extra information
0 is composed of only the PPCM block. Sub-packets in the second and subsequent PPCM access units include a bit stream BS0 except for a directory,
The sub-stream BS0 at the head of the frame is composed of a CRC and extra information.
It is composed of M blocks (including a frame head sample value).

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. -Data rate: "0" in the case of VBR (identifier indicating that the data in the subpacket is compressed data)-Sampling frequency fs and number of quantization bits Qb-Channel allocation information The restart header is a channel for each frame. Correlation circuit A
(Having an addition circuit and a subtraction circuit). The variable rate bit stream data of the format shown in FIG.
The original 2 channel audio signal is decoded by the decoders 3D1 and 3D2 having a configuration of 1 or less.

FIG. 14 is a block diagram showing a second embodiment of the speech encoding apparatus and speech decoding apparatus according to the present invention. The channel correlation circuit A-1 shown in FIG.
a and a subtraction circuit 1b. The addition circuit 1a is a stereo 2
The sum signal (L + R) of the ch signals L and R is calculated, the sum signal (L + R) is divided by 1/2 by the divider 5a, and the divided signal is output to the lossless encoder 2D, and the subtraction circuit 1b
Calculates the difference signal (LR), divides the difference signal (LR) by に よ り by the divider 5b, and outputs the result to the lossless encoder 2D. Lossless encoder 2
D multiplexes these using 1/2 (L + R) and 1/2 (LR) to create a multiplexed signal 250. The multiplexed signal 250 is decoded by the lossless decoder 3D to obtain the original 1/2 (L + R) and 1/2 (LR), which are added to the addition circuit 4a constituting the channel correlation circuit B-1. And a subtraction circuit 4b, respectively, to obtain stereo 2-channel L and R signals as output signals. Note that a series of operations in the lossless encoder 2D and the lossless encoder 2D, such as calculation of a difference, calculation of a predicted value, selection of a minimum prediction residual, and calculation of a predicted value using the minimum prediction residual are performed in the first step. This is performed in the same manner as in the embodiment.
The variable rate bit stream data of the format shown in FIG. 13 is used to identify whether the channel correlation circuit shown in FIG. 1 or the channel correlation circuit shown in FIG. 14 is used, for example, by using an identifier stored in the restart header of the PPCM access unit. Therefore, any one of them can be reliably decoded. Although lossless compression for each frame has been described as an example, the section may be of variable length instead of being fixed.

[0030]

As described above, according to the present invention, in particular, the first and second audio signals of the same sampling frequency are mixed and converted into mutually correlated channels.
The first sample value is obtained from the two correlation signals in response to the audio signal input for each channel, and the prediction residual among a plurality of prediction values of the current signal predicted from the past signal in the time domain is obtained. Since the lossless compression is performed by the minimum linear prediction method, the compression ratio can be improved when predictive coding of the audio signal is performed.

[Brief description of 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 thereto.

FIG. 2 is a block diagram showing the encoder of FIG. 1 in detail.

FIG. 3 is an explanatory diagram showing a format of one frame multiplexed by the multiplexer of FIG. 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.

FIG. 6 is a block diagram illustrating the decoder of FIG. 1 in detail;

FIG. 7 is a block diagram illustrating an encoder according to a second embodiment.

FIG. 8 is a block diagram illustrating a decoder according to a second embodiment.

FIG. 9 is a block diagram illustrating an encoder according to a third embodiment.

FIG. 10 is a block diagram illustrating a decoder according to a third embodiment.

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

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

FIG. 13 is a format explanatory diagram showing an aspect different from FIG. 3 of the audio (A) packet of the compressed PCM (PPCM) shown in FIG. 5;

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

[Explanation of symbols]

1a, 4a Addition circuit (addition means) 1b, 4b Subtraction circuit (subtraction means) 5a, 5b Divider 11D1 Difference calculation circuit (first difference calculation means) 11D2 Difference calculation circuit (second difference calculation means) 12a-1 -12a-n predictor (subtractors 13a-1
13a-n and the buffer / selector 16D1 constitute a first predictive encoding means. ) 12b-1 to 12b-n predictors (subtractors 13b-1 to 13b-1)
13b-n and the buffer / selector 16D2 constitute a second predictive encoding means. 13a-1 to 13a-n, 13b-1 to 13b-n Subtractors 16D1, 16D2, 16A, 16S, 16L, 16R
Buffer / selector 15A Prediction circuit (third with buffer / selector 16A)
Of predictive encoding means. ) 15S prediction circuit (4th with buffer / selector 16S)
Of predictive encoding means. ) 15L prediction circuit (fifth with buffer / selector 16L)
Of predictive encoding means. ) 15R prediction circuit (6th with buffer / selector 16R)
Of predictive encoding means. )

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G10L 9/14 J 9/18 D

Claims (1)

[Claims] A step of mixing first and second audio signals having the same sampling frequency and converting the mixed signals into channels correlated with each other; and converting the two correlated signals converted by the step into channels, A linear prediction method that obtains a leading sample value in response to an input audio signal and that has a minimum prediction residual among a plurality of prediction values of a current signal predicted from a past signal in the time domain. Selecting and predictively encoding; a compressed PCM having header information and user data and having a compressed PCM access unit in the user data.
Storing in the subpacket in the compressed PCM access unit in the data format the predicted data including the head sample value selected in the step, the prediction residual, and the linear prediction method.