JPH0632032B2 - Speech band signal coding method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial Field of Application> The present invention sets a low bit rate waveform coding method for a voice band signal (voice signal, data modem signal, etc.), and particularly sets a transmission information amount to 16 kbit / sec or less. Such an encoding method and device.

<Prior Art and Problems Thereof> A multi-pulse drive type audio encoding system has recently been proposed as a system for encoding an audio signal with a transmission information amount of about 16 kbit / sec or less. This is a system in which a plurality of pulse sequences (multi-pulses) representing a driving sound source signal sequence are transmitted at a short time by the encoder side at A-B-S (ANALYSIS-BY-SYNTHE).
This is a method of coding and transmitting this pulse sequence sequentially using the method of SIS). The present invention relates to this system. For more details on this method, see ICA by BSATAL et al.
Proceedings of ICSP, 1982 614-617
"A. New Model of L.P.
C. Exhibition. Four. Producing. natural. Sounding. speech. at.
Row. bit. Rates "(" A NEW MODEL OF LPC EXCIT
ATION FOR PRODUCING NATURAL-SOUNDING SPEECH AT LOW
Since it is explained in the paper (reference 1) entitled "BIT RATES"), only a brief explanation will be given here.

FIG. 1 is a block diagram showing the processing on the encoder side in the conventional method described in Document 1 above. In the figure,
Reference numeral 100 denotes an encoder input terminal to which the A / D converted audio signal sequence x (n) is input. Reference numeral 110 denotes a buffer memory circuit, which stores the audio signal sequence for one frame (for example, 80 samples when the frame length is 10 msec in the case of 8 KHZ sampling). The output value of the buffer memory circuit 110 is output to the subtractor 120 and the K parameter calculation circuit 180. However, according to Ref. 1, instead of the K parameter, the Reflection COE
FFICIENTS) is the same parameter as the K parameter. K parameter calculation circuit 18
0 uses the output value of the buffer memory circuit 110, calculates the K parameter K _i representing the audio signal spectrum for each frame for the 16th order (1 ≦ i ≦ 16) according to the covariance method, and outputs them to the synthesis filter circuit 130. Output. 140 is
A sound source pulse generation circuit that generates a predetermined number of pulse sequences in one frame. here,
This pulse sequence will be referred to as d (n). Source pulse generation circuit 1
An example of the sound source pulse sequence generated by the 40 is shown in FIG. In FIG. 2, the horizontal axis represents discrete time and the vertical axis represents amplitude. Here, the case where eight pulses are generated in one frame is shown. The pulse sequence d (n) generated by the sound source pulse generation circuit 140 drives the synthesis filter 130. The synthesis filter 130 is
Input d (n), and play signal corresponding to audio signal x (n) And outputs it to the subtractor 120. Here, the synthesis filter 130 inputs the K parameter K _i , converts the K parameter K _i into these prediction parameters ai (1 ≦ i ≦ 16), and calculates the reproduction signal x (n) using ai.

Can be expressed as follows using d (n) and ai.

In the above equation, P indicates the order of the synthesis filter, where P = 1.
6 is set. The subtractor 120 outputs the original signal And the reproduction signal x (n) and the difference e (n) are calculated, and the weighting circuit 190
Output to. 190 inputs e (n) and weighting function w
Using (n), the weighting error e _w (n) is calculated according to the following equation.

e _w (n) = w (n) * e (n)-(2) In the above equation, the symbol "*" represents convolution integral. Also,
The weighting function w (n) is used to perform weighting on the frequency axis, and when its Z-transformed value is W (Z), it is expressed by the following equation using the prediction parameter a _i of the synthesis filter.

In the above equation, r is a constant of 0 ≦ r ≦ 1, and determines the frequency characteristic of W (Z). That is, when r = 1, W (Z) = 1 and the frequency characteristic of W (Z) is flat. On the other hand, when r = 0, W (Z) has the inverse characteristic of the frequency characteristic of the synthesis filter. Therefore, the characteristic of W (Z) can be changed by the value of r. Also, the reason why the W (Z) characteristic is determined depending on the frequency characteristic of the synthesizing filter as shown in the equation (3) is that the perceptual masking effect is used. That is,
At a place where the power of the spectrum of the input audio signal is large (for example, near the formant frequency), even if the error with the spectrum of the reproduction signal is a little large, the error is hard to hear, which is due to the auditory property. FIG. 3 shows the spectrum of the input audio signal in a certain frame and an example of the frequency characteristic of W (Z). Here, r = 0.8. In the figure, the horizontal axis represents frequency (up to 4 KHz) and the vertical axis represents logarithmic amplitude (up to 60 dB). The upper curve represents the spectrum of the audio signal, and the lower curve represents the frequency characteristic of the weighting function.

Returning to FIG. 1, the weighting error e _w (n) is fed back to the error minimization circuit 150. Error minimization circuit 15
For 0, the value of e _w (n) for one frame is stored, and using these, the weighted error power ε is calculated according to the following equation.

Here, N represents a sample for calculating the error power. In the method of Reference 1, this time length is set to 5 msec, which corresponds to N = 40 in the case of 8 KHz sampling.
Next, the error minimization circuit 150 obtains the amplitude and position of the sound source pulse so as to reduce the error power ε calculated by the equation (4), and outputs this amplitude information and position information to the sound source pulse generation circuit 140. To do. The sound source pulse generation circuit 140 generates a sound source pulse sequence based on this information.

The synthesis filter circuit 130 uses this sound source pulse sequence as a driving source Ask for. In the subtractor 120, the reproduction signal obtained as described above from the error e (n) between the original signal and the reproduction signal previously calculated Is subtracted to obtain a new error e (n). The weighting circuit 190 inputs e (n), calculates a weighting error e _w (n), and feeds it back to the error minimizing circuit 150. The error minimization circuit 150 calculates the error power again,
The amplitude and position of the sound source pulse sequence are adjusted so as to reduce this error power. In this way, a series of processes from generation of the sound source pulse sequence to adjustment of the sound source pulse sequence by error minimization is repeated until the number of pulses in the sound source pulse sequence frame reaches a predetermined number, and the sound source pulse sequence is determined. It

This is the end of the description of the conventional method.

In the case of this method, the information to be transmitted is the K parameter K _i (1 ≦ i ≦ 16) of the synthesizing filter, the pulse position and the amplitude of the sound source pulse sequence, and is arbitrary depending on the number of pulses in one frame. A transmission rate can be realized. further,
It is considered to be one of the effective methods in which a good reproduction sound quality is obtained for a region where the transmission rate is 16 Kbps to 10 Kbps.

However, this conventional method has a drawback that the amount of calculation is very large. This is to calculate the error and error power between the reproduced signal and the original signal based on the pulse when calculating the position and amplitude of the pulse in the sound source pulse sequence, and feed them back to reduce the error power. This is because the pulse position and amplitude are adjusted. Furthermore, it is caused by repeating the processes from the generation of these pulses to the adjustment of the pulse amplitude and the position by feeding back the error power until the number of pulses reaches a predetermined value.

In the case of a transmission bit rate of 16 kbit / sec or less,
In the unvoiced part of the voice signal, the number of sound source pulses cannot be increased sufficiently according to the conventional method, so that good characteristics cannot be obtained in such a part.

As a recent trend, there is a strong demand for good transmission of a voice band data modem signal of about 2400 bits / sec at a transmission bit rate of about 16 kbits / sec. According to the conventional method, it is difficult to obtain good characteristics for a voice band data modem signal because the number of pulses is not sufficiently large.

<Object of the Invention> The object of the present invention is 16 kbit / sec, or 16 kbit / sec.
Voice band signal coding method that can obtain good characteristics with a relatively small amount of calculation not only for voice signals with a transmission bit rate of less than a second but also for voice band data modem signals of about 2400 bits / sec. And to provide the device.

<Configuration of Invention> According to the present invention, on the transmission side, a discrete voice band signal sequence is input, a spectrum parameter sequence representing a short-time spectrum envelope is extracted, and the voice band signal sequence and the spectrum parameter sequence are also extracted. To search for a pulse sequence that can satisfactorily represent the voice band signal sequence, and to create a discrimination code that determines the number of transmitted pulse sequences based on the spectrum parameter sequence extraction result or the pulse sequence search result,
The transmission pulse sequence and the spectrum parameter sequence are encoded according to the discrimination code and output in combination with the discrimination code, and on the receiving side, the discrimination code is separated from the combined code, and the spectrum is determined according to the discrimination code. A code representing a parameter sequence and a code representing the transmission pulse sequence are separated and decoded, and the voice band signal sequence is reproduced using the decoded spectrum parameter sequence and the decoded pulse sequence. A featured voice band signaling method is obtained.

Further, according to the present invention, a parameter calculation circuit that inputs a discrete voice band signal sequence and extracts a spectrum parameter sequence representing a short-time spectrum envelope from the voice band signal sequence, the voice band signal sequence, and the spectrum A pulse sequence search circuit that searches for a pulse sequence that can satisfactorily represent the voice band signal sequence based on a parameter sequence, and determines the number of transmission pulse sequences based on the spectrum parameter sequence extraction result or the pulse sequence search result. A voice band signal sequence encoding device comprising: a discriminating circuit for producing a discriminating code; and means for encoding the transmission pulse sequence and the spectrum parameter sequence in accordance with the discriminating code and outputting in combination with the discriminating code. To be

Further, according to the present invention, a spectrum parameter sequence representing a short-time spectrum envelope is extracted from a discrete voice band signal sequence from the transmitting side, and the voice band signal sequence is based on the voice band signal sequence and the spectrum parameter sequence. Is searched for a pulse sequence that can be expressed well, and a discrimination code that determines the number of transmitted pulse sequences based on the spectral parameter sequence extraction result or the pulse sequence search result is created,
A code output by combining the transmission pulse sequence and the spectrum parameter sequence according to the discrimination code and combining with the discrimination code is input, and the discrimination code is separated from the combined code sequence, and further the spectrum is determined according to the discrimination code. Means for separating and decoding the code representing the parameter sequence and the code representing the pulse sequence, a pulse sequence generation circuit for generating a drive pulse sequence using the decoded pulse sequence, the decoded spectrum parameter sequence and the A voice band signal decoding device is provided which has a synthesis filter circuit for reproducing and outputting a voice band signal sequence using a drive pulse sequence.

<Embodiment> The configuration of the audio encoding system according to the present invention will be described in detail with reference to the drawings. FIG. 4 (a) is a block diagram showing an embodiment of the encoder side of the audio encoding system according to the present invention.
FIG. 6B is a block diagram showing an embodiment of the decoder side.
In FIG. 4 (a), the audio signal sequence x (n) is input terminal 1
The data is input from 95, divided into a predetermined number of samples, and stored in the buffer memory circuit 340.
Next, the K parameter calculation circuit 280 inputs a predetermined number of samples of the audio signal stored in the buffer memory circuit 340, and calculates the LPC parameter representing the spectral envelope of the input audio signal. LPC
There are various parameters, but in the following description, the K parameter is used. The K parameter is the same parameter as the Percoll coefficient. The autocorrelation method and the covariance method are well known as typical methods for calculating the K parameter. Here, the calculation method of the K parameter by the autocorrelation method is described by JOHN M.
AKHOUL) et al. IE TRANSACTIONS ON AS TRANS
ACTIONS ON ASSP) June 1975 issue. See Quantization Properties of Pages 309-321.
Transmission Parameters Linear,
Predictive system "(" QUANTIZATION PROPE
RTIES OF TRANS MISSION PARAMETERS IN LINEAR PREDIC
The method described in the paper (Reference 2) entitled "TIVE SYSTEMS") is cited below.

E ₀ = R (o) (5a) a _i ⁽ⁱ⁾ = k _i (5c) ▲ a _j ⁽ⁱ⁾ ▼ = ▲ a _j ^(i-1) ▼ + ▲ k _i a _ij ^(i-1) ▼, (1 ≤
j ≦ i−1) (5d) E _i = (1-k _i ² ) · E _i-1 (5e) a _j = a _j ^(p) , (1 ≦ j ≦ p) (5f) Formula (5a) Therefore, equation (5f) can be recursively solved with j = 1, 2, ... P. In the equation, k _i indicates the i-th order K parameter value. R (i) is the delay time i with respect to the input voice.
The autocorrelation number of is shown. P indicates the prediction analysis order. ▲ a
_j ^(p) ▼ indicates the j-th linear prediction coefficient in the case of the analysis order d. Here, the value of E _i in equation (5e) represents the prediction error power in the prediction of order i. Therefore, the prediction error power of the prediction of order i can be monitored at each stage of the calculation. The normalized prediction error can be expressed as follows using E _i .

V _i = E _i / R (o) (6) When i = p, use equation (5e) Can be expressed as Here, 1 / Vp is also called a prediction gain. Therefore, if the equation (7) is used, the normalized prediction error in the p-th order prediction analysis can be known. This is the end of the description of the K parameter calculation method based on the autocorrelation method.

Returning to FIG. 4 (a), the K parameter calculation circuit 280
Predetermined order M ₁ according to equation (5a) to equation (5e)
The K parameter K _i (1 ≦ i ≦ M ₁ ) of (for example, M ₁ = 4) is calculated. Also, according to Eq. (7), the M ₁ -order normalized prediction error V _M1
To calculate. Next, the obtained normalized prediction error V _M1 is compared with a predetermined threshold value, and if V _M1 is smaller than the threshold value, the input signal is determined to be voiced as an example.
On the other hand, if V _M1 is larger than the threshold value, the input voice is determined to be unvoiced. The reason for this is that in the case of a voice signal, the correlation is large in the voiced part, so that it is easy to predict and the normalized prediction error is a considerably small value. On the other hand, the unvoiced part of the voice signal and the data modem signal are difficult to predict because the correlation is small, and the reason is that the normalized prediction error is not so small. However, for simplification of the description, the voiced and unvoiced voices are classified into two types, but it is not necessary to specifically classify into voiced voices and unvoiced voices, and two or more types may be used. The K parameter calculation circuit 280 sets the K parameter coding circuit 200 and the impulse response calculation circuit 2 as the voiced / unvoiced discrimination result using the normalized prediction error V _P as 1-bit information d.
10, pulse calculation circuit 390, synthesis filter circuit 400
To the weighting circuit 410, the encoding circuit 470, and the multiplexer 450. Furthermore, K parameter calculation circuit 2
80, K parameter value _{K i (1 ≦ i ≦ M} 1, for example, M ₁ = 4) obtained up to the _primary M when discrimination result is silent with K
It is output to the parameter encoding circuit 200. In this case, since the correlation of the signals is small, the improvement of the prediction gain is very small even if M ₁ is set to the fourth order or more. On the other hand, when the discrimination result is voiced, in order to more accurately represent the spectral envelope of the speech signal, K parameter values K _i (1) up to M _{2 nd} order (M ₂ ≧ M ₁ , for example M ₂ = 12) ≦ i ≦ M ₂ ) is continuously calculated,
The K _i (1 ≦ i ≦ M ₂ ) is output to the K parameter encoding circuit 200.

The K parameter encoding circuit 200 receives the voiced / unvoiced discrimination information d and the K parameter value K _i from the K parameter calculation circuit 280.
Enter and. The K parameter coding circuit 200 has two kinds of quantization characteristics, that is, an optimum quantization characteristic for voiced voice and an optimum quantization characteristic for unvoiced voice. These characteristics are switched according to the discrimination information d, and the input K parameter _Ki is encoded. , Code l _ki is output to the multiplexer 450. In addition, the K parameter encoding circuit 200 uses the K parameter decoded value K _i obtained by decoding l _ki, as described in (5c), (5d), and (5f) above.
The prediction coefficient value a ′ _i is converted using the formula. Voice /
The order p is switched to M ₁ or M ₂ using the unvoiced discrimination information d. The K parameter encoding circuit 200 uses the prediction coefficient value
The a′K _i is output to the impulse response calculation circuit 210, the weighting circuit 410, and the synthesis filter circuit 400.

Next, the impulse response calculation circuit 210 inputs the voiced / unvoiced discrimination information d from the K parameter calculation circuit 280 and the prediction coefficient value a ′ _i from the K parameter encoding circuit 200, and outputs the weighted synthesis filter of the following equation. The impulse response h _w (n) representing the transfer function is calculated by a predetermined number of samples.

Where P represents the order of the prediction count a _'i. P is switched according to the voiced / unvoiced discrimination information d. In the case of voiced, P is
M ₂ (eg 12) Set next, P is M ₁ if unvoiced
(Eg 4) is set next. Also, W (Z) is the above (3)
It is a Z-transform expression of the weighting function shown by a formula. However, the order P is switched to M ₂ or M ₁ according to the voiced / unvoiced information d. The impulse response calculation circuit 210 is an impulse response
It outputs h _w (n) to the autocorrelation coefficient calculation circuit 360 and the cross-correlation coefficient calculation circuit 350.

Next, the autocorrelation coefficient calculation circuit 360 inputs the impulse response h _w (n) from the impulse response calculation circuit 210 and calculates the autocorrelation coefficient R _hh (.) According to the following equation for a predetermined delay time τ. To do.

The autocorrelation factor R _hh (τ) is output to the pulse calculation circuit 390.

Next, the subtractor 285 inputs the audio signal x (n) accumulated in the buffer memory circuit 340, subtracts the output sequence of the synthesis filter circuit 400 by one frame sample from x (n), and subtracts the result e (n ) Is output to the weighting circuit 410.

Next, the weighting circuit 410 outputs the subtraction result from the subtractor 285.
e (n) is input, the prediction coefficient value a ′ _i is input from the K parameter encoding circuit 200, and the K parameter calculation circuit 280 is input.
The voiced / unvoiced discrimination information d is input from the above, weighting is applied to e (n), and e _w (n) is output. Here, e _w (n) is a Z-transform expression and can be written as

E _w (Z) ＝ E (2) ・ W (Z) (10) where E _w (Z) and E (Z) are the Z conversion value of e _w (n) and the Z conversion value of e (n), respectively. Indicates. W (Z) represents the Z-transformed value of the weighting function represented by the above equation (3). However, the degree p of W (Z) is voiced /
It is switched to M ₂ or M ₁ according to the unvoiced information d. The weighting circuit 410 outputs the obtained e _w (n) to the cross correlation coefficient calculation circuit 350.

Next, the cross correlation coefficient calculation circuit 350 uses the weighting circuit 41.
Input e _w (n) from 0, and impulse response calculation circuit 2
The impulse response h _w (n) is input from 10 and the number of cross-correlation _parameters h _x (n) is calculated by a predetermined number of samples according to the following equation.

The cross correlation number _hx ( _.multidot. ) Is output to the pulse calculation circuit 390.

Next, the pulse calculation circuit 390 uses the cross correlation coefficient calculation circuit 3
The cross-correlation _{coefficient hx} (.) _Is input from 50, the auto-correlation coefficient R _hh (.) _Is input from the auto-correlation coefficient calculation circuit 360, and the voiced / unvoiced discrimination information d is input from the K-parameter calculation circuit 280. . Here, the pulse calculation circuit 390 switches the number of pulses required in one frame according to the voiced / unvoiced discrimination information d. That prompted the L ₁ pulses in the case of voiced, in the case of silent determine the L ₂ pulses. However, L ₁ <
Set to L ₂ . The reason why it is necessary to increase the number of pulses in the unvoiced case as compared with the voiced case is that the unvoiced case has a smaller prediction gain than the voiced case as described above. Here, the number of pulses must be determined according to the transmission bit rate. For example, the transmission bit rate is 16 kbit /
In terms of seconds, according to the quantization bit allocation in the quantization circuit described later, L ₁ = 32 for voiced and L ₂ = unvoiced.
It will be about 50 pieces.

The pulse calculation circuit 390 sequentially calculates a pulse sequence that minimizes the weighted error power between the input signal and the combined signal, pulse by pulse, according to the following equation.

Here, g _i represents the amplitude of the i-th pulse in the frame. m _i indicates the sample position within the frame of the i-th pulse. L represents the number of pulses to be obtained in one frame,
This value is L according to the voiced / unvoiced discrimination information as described above.
Switchable to ₁ (if voiced) or L ₂ (if unvoiced). The position m _i of the pulse is obtained from the position in the frame where the maximum absolute value of g _i is taken.

Next, the process of obtaining the pulses one by one according to (12) will be described with reference to the drawings. FIG. 5A shows the cross-correlation count for one frame which is calculated by the cross-correlation count calculation circuit 350 and output to the pulse calculation circuit 390. In the figure, the horizontal axis represents the sample time within one frame. Frame length is 1
It is set to 60. The vertical axis is the amplitude. Figure 5 (b) is (12)
It is a figure which shows the _1st pulse g1 calculated | required according to a formula.
FIG. 5 (c) is a diagram after subtracting the influence of the pulse obtained in FIG. 5 (b). FIG. 5 (d) is a diagram in which the _second pulse g ₂ is obtained. FIG. 5 (e) is a diagram after subtracting the influence of the _second pulse g ₂ . The processes of FIG. 5 (d) to (e) are repeated to obtain L ₁ or L ₂ pulses.

Returning to FIG. 4A, the pulse calculation circuit 390 outputs the pulse sequence obtained according to the equation (12) to the encoding circuit 470.

Next, the encoding circuit 470 inputs the pulse sequence from the pulse calculation circuit 390 and the voiced / unvoiced discrimination information d from the K parameter calculation circuit 280. The encoding circuit 470 is
According to the voiced / unvoiced discrimination information d, the number of quantization bits and the quantization characteristic are switched for voiced and unvoiced. The reason why the quantization characteristics are switched is that the frequency distributions of the pulse amplitudes are different between voiced and unvoiced voices, so that optimum quantization is applied to each distribution. The encoding circuit 470 encodes the amplitude and position of the input pulse, and the multiplexer 450
Output to. Also, the amplitude of the pulse and the decoded value g ′ _{i of} the position,
m ′ _i is output to the pulse generation circuit 420. Here, various pulse sequence encoding methods can be considered. One is a method of separately encoding the amplitude and the position of the pulse sequence, and the other is a method of encoding the amplitude and the position together.

An example of the former method will be described. First, as the encoding method of the amplitude of the pulse sequence, the maximum value of the amplitude of the pulse sequence in the frame is used as a normalization count, and the amplitude of each pulse is normalized using this value, and then quantized and encoded. A method can be considered. As the quantization characteristic, the optimum characteristic according to the amplitude distribution in each case of voiced and unvoiced is used. Further, the amplitude of each pulse may be quantized and encoded after being converted into another parameter having an orthogonal relationship. Also, bit allocation may be changed for each pulse amplitude. Next, various methods can be considered for encoding the pulse position.
For example, a run length code or the like which is well known in facsimile signal coding or the like may be used. This is the code "0"
Alternatively, the length following "1" is represented by using a predetermined code sequence. Further, conventionally well-known logarithmic compression encoding or the like can be used for encoding the normalization coefficient.

An example of quantized bit allocation for voiced and unvoiced cases is shown below. The transmission bit rate is 16 kbit / sec. If the discrimination information d is voiced, the quantization bit number of the pulse amplitude is 5 bits and the bit number of the pulse position is 3 bits. On the other hand, when the discrimination information is unvoiced, the quantization bit number of the pulse amplitude is 4 bits and the bit number of the pulse position is 2 bits. According to this bit allocation, when the transmission bit rate is 16 kbit / sec, the number of pulses for voiced voice is 32 and the number of pulses for unvoiced voice is about 50, as described above.

Regarding the encoding of the pulse sequence, it is needless to say that the best known method can be used without being limited to the encoding method described here.

Returning to 4 (a), the pulse generating circuit 420, the pulse sequence decoded value g _'i, m' with _i m 'to the position of the _i amplitude g' generates a drive pulse sequence with _i. Pulse generation circuit 42
0 outputs the drive pulse sequence to the synthesis filter circuit 400.

The synthesis filter circuit 400 inputs the drive pulse sequence from the pulse generation circuit 420, and receives the K parameter calculation circuit 280.
The voiced / unvoiced discrimination information d is input from, and the prediction coefficient decoded value a ′ _i is input from the K parameter encoding circuit 200. The synthesis filter circuit 400 uses the input drive pulse sequence and decoded prediction coefficient a ′ _i to generate a response signal sequence for one frame. Is calculated according to the following formula.

here The value of is calculated for two frames (1 ≦ n ≦ 2N). d (n)
Represents a drive signal, and when 1 ≦ n ≦ N, the pulse generation circuit 4
The drive pulse sequence input from 20 is used. Also N + 1 ≤
When n ≦ 2N, a sequence of all 0 is used. The order p is the discrimination information d
In the case of voiced, M ₂ (for example, 12th) order is selected, and in the case of unvoiced, M ₁ (for example, 4th) order. Found in (13) Of the second frame Is output to the subtractor 285.

Next, the multiplexer 450 inputs the output code of the encoding circuit 470, the output code of the K parameter encoding circuit 200, and the 1-bit code representing the discrimination information from the K parameter encoding circuit 280, and combines them to the transmitting side. Output from the output terminal 480 to the communication path. This is the end of the description of the encoder side of the speech encoding system according to the present invention.

Next, the decoder side of the voice encoding system according to the present invention will be described with reference to FIG. 4 (b). Demultiplexer 500
Inputs the combined code from the decoder side input terminal 490. Of the input codes, the demultiplexer 500 separates the code representing the K parameter, the code representing the pulse sequence, and the 1-bit code representing the voiced / unvoiced discrimination information, and outputs the code representing the K parameter to the K parameter decoding circuit 520. Then, the code representing the pulse sequence is output to pulse sequence decoding circuit 530, and the 1-bit code representing the voiced / unvoiced discrimination information is output to K parameter decoding circuit 520, pulse sequence decoding circuit 530, and synthesis filter circuit 550.

Next, the pulse sequence decoding circuit 530 inputs the code representing the voiced / unvoiced discrimination information and the code representing the pulse sequence, and according to the code representing the voiced / unvoiced discrimination information, L ₁ (for example, 32) Decode the pulse sequence of. on the other hand,
When unvoiced, L ₂ (eg 50) pulse sequences are decoded. The amplitude and position information of the decoded pulse sequence are output to the pulse generation circuit 540. Pulse generation circuit 540
Receives the decoded amplitude and position information, generates a drive pulse sequence, and outputs the drive pulse sequence to the synthesis filter circuit 550.

Next, the K parameter decoding circuit 520 inputs a code representing the voiced / unvoiced discrimination information and a code representing the K parameter, and in accordance with the code representing the voiced / unvoiced discrimination information, in the case of voiced M ₂ (for example, 12) th order Decode the K parameters of On the other hand, in the case of being unvoiced, the M ₁ (for example, 4) th order K parameter is decoded. The decoded and obtained parameter value K _i is output to the synthesis filter circuit 550.

Next, the synthesis filter circuit 550 outputs a code representing the voiced / unvoiced discrimination information, the driving pulse sequence, and the K parameter decoded value K _i.
Enter and. The K parameter decoded value K _i is the same as (5c), (5
Prediction coefficient value a ′ _i is converted using d) and (5f). At this time, the order p is determined according to the code representing the voiced / unvoiced discrimination information.
Switch to M ₁ or M ₂ . Synthesis filter circuit 550
Is the combined signal according to Is calculated for one frame and is output from the reception side output terminal 560.

Here, d (n) represents a drive pulse sequence. The order p is switched to M ₁ or M ₂ according to a code representing voiced / unvoiced discrimination information. This is the end of the description on the decoder side according to the present invention.

According to the configuration of the present embodiment, since the pulse sequence is obtained according to the above-mentioned equation (12), the reproduction signal is obtained by driving the synthesis filter with the sound source pulse as in the conventional method of Document 1 to obtain the original signal. Since there is no path for adjusting the pulse by hooding back the squared error of 1 and there is no need to repeat the processing, the amount of calculation can be greatly reduced. However, the pulse calculation algorithm is not limited to the method described in the embodiment, and if the calculation amount is allowed to increase, a pulse calculation algorithm based on the A-B-S method as shown in Reference 1 is used. Good.

In the pulse calculation method shown in the equation (12), pulses were calculated one by one. In this method, the amplitudes of a plurality of pulses obtained in the past may be readjusted when the next pulse is calculated. By doing this, the characteristics are improved when the distance between the pulses is short and the pulses are not independent of each other. As a method for obtaining the sound source pulse, another good pulse sequence calculation method such as a method for calculating a more optimal pulse sequence may be used.

Further, in the present embodiment, the normalized prediction error is calculated on the encoder side in accordance with the above equation (7), and the voiced / unvoiced discrimination information is created according to this value. May be as follows. Now, assume that the transmission bit rate is 16 kbit / sec. Pulse calculation circuit 3
At 90, the number L ₁ (for example, 50) of pulses when it is determined to be unvoiced is obtained, and at the encoding circuit 470, for example, 4-bit quantization is performed on the amplitude of each pulse, and each pulse position is a 2-bit code. Express with. The amplitude and position of each pulse are decoded, and the error power E ₁ is calculated according to the following equation.

Here, R _ee (o) is N of the output value e _w (n) of the weighting circuit 410.
The power of the sample is shown. L is the number of pulses (in this case L _1), g _'i is the i th decoded pulse amplitude of the pulse, m' _i is the i th decoded position of the pulse, _hx (·)
Indicates the number of cross correlations. Further, among the L ₁ pulses, the number L ₂ (for example, 32) pulses when the voices are judged to be voiced in order from the largest amplitude are selected, and the encoding circuit 470 performs 5-bit quantization on each pulse amplitude. , Each pulse position is represented by a 3-bit code and decoded. Using the decoded value, the error power E ₂ is calculated according to the above equation (15). However, (1
L in equation (5) must be L ₂ . Next, E ₁ and E ₂ are compared, and if E ₁ is smaller, it is determined to be unvoiced, the discrimination code is set to a code indicating unvoiced, and the number of pulses is L ₁ . On the other hand, if E ₂ is smaller, it is determined to be voiced, the discrimination code is set to a code indicating voiced, and the number of pulses is L ₂ . With such a configuration, it is possible to perform voiced / unvoiced discrimination based on the overall characteristic including the quantization effect, so that the characteristic is further improved.

Further, in the present embodiment, using the voiced / unvoiced discrimination information, the encoder side switches the K parameter encoding circuit 200, the quantization characteristic of the encoding circuit 470, and the quantization bit distribution, and the decoder side performs the K parameter decoding. The decoding characteristics of the circuit 520 and the pulse decoding circuit were switched. In order to further simplify the device configuration, the quantization characteristic, the quantization bit allocation, and the decoding characteristic may be voiced or unvoiced and may be the same characteristic without switching.

Further, in this embodiment, the order of the K parameter is switched by the K parameter calculation circuit 280 on the encoder side using the voiced / unvoiced discrimination information. On the other hand, K on the decoder side
Although the orders of the parameter decoding circuit 520 and the synthesis filter circuit 550 are switched, the switching operation regarding this order may not be performed.

Further, in the present embodiment, the order of the synthesis filter circuit 550 is switched by inputting the voiced / unvoiced discrimination information, but the switching operation using the voiced / unvoiced discrimination information may be omitted. This is because the order of the K parameter decoded value input from the K parameter decoding circuit 520 has already been switched according to the voiced / unvoiced discrimination information.

Further, in the present embodiment, the pulse calculation circuit 390 uses the voiced / unvoiced discrimination information to switch the number of pulses L to be obtained in the frame, but the number of pulses calculated by the pulse calculation circuit 390 is the same for both voiced and unvoiced L. ₁ (eg 5
0) pieces may be calculated and the number of pulses to be transmitted may be switched using voiced / unvoiced discrimination information when transmitting a code representing a pulse sequence in the multiplexer 450. With such a configuration, when switching to the one with a smaller number of pulses and transmitting, for example, from the one with a large pulse amplitude
L ₂ (for example, 32) pieces may be selected and transmitted.

In addition, in this embodiment, the type of switching the pulse number is
Although the number of pulses is L ₁ or L ₂ , the number of pulses may be switched to three or more. However, in this case, it is necessary to prepare two or more thresholds for the voiced / unvoiced discrimination on the encoder side and increase the number of bits of the discrimination code transmitted to the decoder side.

In the configuration of the present embodiment, when calculating the autocorrelation coefficient of the impulse response sequence representing the short-time spectrum structure, the impulse response calculation circuit 210 calculates the impulse response using the K parameter decoded value, and then the impulse response is calculated. The autocorrelation coefficient calculation circuit 360 calculates the autocorrelation coefficient using the response. As is well known in the field of digital signal processing, the autocorrelation number of the impulse response corresponds to the power spectrum. Therefore, first, the power spectrum is obtained using the K parameter decoded value,
After that, the correspondence relationship may be used to calculate the autocorrelation number. On the other hand, when calculating the correlation coefficient between the speech signal and the impulse response representing the short-time spectrum envelope, the output value e _w (n) of the weighting circuit 410 and the K parameter decoder K _i are used in the configuration of this embodiment. Impulse response h _w (n) calculated by the impulse response calculation circuit 210
_Was used to calculate the cross-correlation number _hx (•). As is well known, the cross correlation number corresponds to the cross power spectrum. Therefore, the configuration may be such that the cross-power spectrum is first obtained using e _w (n) and K _i, and then the number of cross-correlation is calculated. Regarding the correspondence between the power spectrum and the autocorrelation number and the correspondence between the cross power spectrum and the crosscorrelation number, "Digital Signal Processing" by AVOPPENHEIM et al. SIGNAL PROCESSING ")
Since it has been described in detail in Chapter 8 of the book entitled (Reference 3), its explanation is omitted here.

In the present embodiment, the encoding of the pulse sequence within one frame is performed by the encoding circuit 470 of FIG. 4 (a) after all the pulse sequences have been obtained. Included in the calculation, every time one pulse is calculated,
The encoding may be performed and the next pulse may be calculated. By adopting such a configuration, a pulse sequence that minimizes an error including coding distortion can be obtained, so that the quality can be further improved.

According to this embodiment, there is almost no reproduced signal near the frame boundary due to the discontinuity of the waveform at the frame boundary.
This is because when the encoder side calculates the pulse sequence of the current frame, the response signal sequence obtained by driving the synthesis filter by the drive excitation pulse sequence of one frame past is extended to the current frame and obtained. This is due to the fact that the pulse sequence of the current frame is calculated for the result of subtracting this from the input audio signal system. Further, although the case where the frame length is constant has been described in the present embodiment, a variable length frame in which the frame length is temporally changed may be used. Further, the number of sound source pulses generated in one frame may not be constant. For example, the number of pulse sequences in each frame may be changed so that the S / N is constant.

Further, in this embodiment, the K parameter is used as the parameter representing the spectrum envelope of the short-time speech signal sequence, but other well-known parameters (for example, LSP parameter etc.) may be used. Further above
The weighting function W (Z) may be omitted in Eqs. (8) and (10).

Further, in the present embodiment, in order to prevent the quality deterioration due to the discontinuity of the reproduced waveform at the frame boundary, the response signal sequence derived from the driving sound source pulse one frame before the current frame is calculated, and the current frame is input. After subtracting this response signal from the voice, the pulse sequence was calculated. As shown in FIG. 6, the data used to calculate the pulse sequence was:
It may be configured so as to include data of a frame transmitting a pulse and data in the past. In FIG. 6, N _T indicates a frame for transmitting a pulse, and N indicates a frame for calculating a sound source pulse. With such a configuration, it is not necessary to calculate the response signal sequence derived from the driving sound source pulse of one frame past.

<Advantages of the Invention> As described above, according to the present invention, the number of pulses per frame is changed so that a reproduction signal of good quality can be provided at all times, so that the transmission bit rate is about 16 kbit / sec. It was difficult to obtain good characteristics when the number of pulses was not enough in 2. Not only could the characteristics of the consonant part of the audio signal be improved, but it was also difficult to obtain good characteristics 2400 bits There is an effect that a voice band data modem signal of about 1 / second can be satisfactorily transmitted.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a conventional system, FIG. 2 is a diagram showing an example of a sound source pulse sequence, and FIG. 3 is a frequency characteristic of an input audio signal sequence and a frequency characteristic of a weighting circuit shown in FIG. FIG. 4 (a) and FIG. 4 (b) are block diagrams showing an embodiment of a voice coding system according to the present invention, and FIG. 5 (a).
(E) is a figure which shows an example of a pulse search process, FIG. 6 is a figure for demonstrating the positional relationship between a pulse transmission frame and a sound source pulse calculation frame. In the figure, 110,340 ... buffer memory circuit, 120,285
...... Subtraction circuit, 130,400,550 …… Synthesis filter circuit, 14
0,420,540 …… Pulse generation circuit, 150 …… Error minimization circuit, 180,280 …… K parameter calculation circuit, 190,410 …… Weighting circuit, 200 …… K parameter coding circuit, 210 ……
Impulse response calculation circuit, 350 ... Cross-correlation function calculation circuit, 360 ... Autocorrelation function calculation circuit, 390 ... Pulse calculation circuit, 470 ... Encoding circuit, 450 ... Multiplexer, 50
0 ... Demultiplexer, 520 ... K parameter decoding circuit, 530 ... Pulse decoding circuit, respectively.

Claims (3)

[Claims] 1. A transmitter side inputs a discrete voice band signal sequence, extracts a spectrum parameter sequence representing a short-time spectrum envelope, and outputs the voice band signal based on the voice band signal sequence and the spectrum parameter sequence. A pulse sequence that can represent a sequence satisfactorily is searched for, a discrimination code that determines the number of transmission pulse sequences based on the spectral parameter sequence extraction result or the pulse sequence search result is created, and the transmission pulse sequence and the The spectrum parameter sequence is encoded and output in combination with the discrimination code, the receiving side separates the discrimination code from the combined code, and the code representing the spectrum parameter sequence according to the discrimination code and the transmission pulse sequence. Is separated and decoded, and the decoded spectrum parameter sequence and Voice band signaling method characterized by using a serial decoded pulse sequence and to reproduce the voice band signal sequence. 2. A parameter calculation circuit for inputting a discrete voice band signal sequence and extracting a spectrum parameter sequence representing a short-time spectrum envelope from the voice band signal sequence, the voice band signal sequence and the spectrum parameter sequence. Based on a pulse sequence search circuit that searches for a pulse sequence that can satisfactorily represent the voice band signal sequence, and a discrimination code that determines the number of transmission pulse sequences based on the spectrum parameter sequence extraction result or the pulse sequence search result. A voice band signal sequence encoding apparatus comprising: a discriminating circuit to be produced, and means for encoding the transmission pulse sequence and the spectrum parameter sequence in accordance with the discriminating code and outputting in combination with the discriminating code. 3. A spectrum parameter sequence representing a short-time spectrum envelope is extracted from a discrete voice band signal sequence from the transmitting side, and the voice band signal sequence is made good based on the voice band signal sequence and the spectrum parameter sequence. A pulse sequence that can be represented by, and create a discrimination code that determines the number of transmission pulse sequences based on the spectrum parameter sequence extraction result or the pulse sequence search result, and according to the discrimination code, the transmission pulse sequence and the spectrum parameter sequence And a code output by combining with the discrimination code is input, the discrimination code is separated from the combined code sequence, and a code representing a spectrum parameter sequence and a code representing a pulse sequence are further generated according to the discrimination code. A means for separating and decoding, and a drive pattern using the decoded pulse sequence. A voice band signal including a pulse sequence generation circuit for generating a voice sequence, and a synthesis filter circuit for reproducing and outputting a voice band signal sequence using the decoded spectrum parameter sequence and the driving pulse sequence. Decoding device.