JPH0632034B2 - Speech coding method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a low bit rate waveform coding method for a voice signal, and more particularly to a coding method that reduces the amount of transmission information to about 16 kbit / sec or less.

(Prior Art) As an effective method for encoding a voice signal with a transmission information amount of about 16 kbit / sec or less, an error between a signal reproduced from a drive source signal sequence of the voice signal and an input signal is used. A method is known in which a search is performed every short time under the condition of the minimum. B.S.Atal (B.
S.ATAL) and others represent the driving sound source signal sequence with multiple pulses, and the amplitude and phase are analyzed at the encoder side every short time by analysis-by-synthesis.
s); The method obtained by the AbS method is effective. The explanation for this is the 1982 ICSAS
PIC (ICASSP) 's Proceedings 614-617, "A new model of LPC Excitement for Producing Natural Sounding Speech at Low Bit Rate (A new mo
del of LPC excitation for producing natural soundi
ng speech at low bit rates) ”(reference 1), so detailed description is omitted here. Since the conventional method of Document 1 uses the A-B-S method as a means for obtaining a pulse sequence, it has a drawback that the amount of calculation is very large.
On the other hand, in Japanese Patent Application No. 57-231603, "Voice coding method" (reference 2), a method is proposed in which the amount of calculation is greatly reduced in order to obtain the pulse sequence. With these methods, the transmission rate is 16 kbit /
It has been reported that good playback sound quality can be obtained in a region of less than a second.

Here, the conventional method of Document 2 will be briefly described. A driving sound source sequence consisting of K pulse sequences in one frame is expressed as follows.

Here, δ (·) is δ of Kronecker. N represents the frame length, and g _k represents the amplitude of the pulse standing at the position l _k . Playback signal obtained by inputting d (n) to the synthesis filter Is the prediction coefficient of the synthesis filter α _i (i = 1, ..., M,
If M is the order of the synthesis filter), it can be written as follows.

Input voice signal x (n) and re-output signal The weighted squared error J in one frame with Becomes Where * is a symbol indicating convolution,
w (n) represents a weighting function. The weighting function is introduced in order to minimize the auditory error between the input audio signal and the reproduced signal. According to the auditory masking effect, noise is suppressed in the band where the voice energy is high. The weighting function weights the error in consideration of such auditory characteristics. As the weighting function, the Z-transform W (Z) is calculated from the prediction parameter α _i of the synthesis filter and the real constant γ that satisfies 0 ≦ γ ≦ 1. What is represented is proposed (Reference 1). further Z conversion of Then, Eq. (3) is expressed as follows.

Also, from the relationship of equation (2) Is as follows.

here H (Z) is the Z conversion of the synthesis filter, and D (Z) is the Z conversion of the driving sound source. Substituting (5) into (4), J = | X (Z) W (Z) -H (Z) W (Z) D (Z) | ² − (66).

Therefore, the inverse Z-transformed signals of X (Z) W (Z) and H (Z) W (Z) are x _w (n) = x (n) * w (n) and h _w (n) = h, respectively. When written as (n) * w (n), (6)
Is as follows.

Source pulse sequence amplitude g _k position l that minimizes Eq. (7)
To obtain _k , the expression (7) is partially differentiated by g _k and is set to 0,
That is Take advantage of the relationship.

Where ψ _xh (・) is the cross-correlation function sequence calculated from x _w (n) and h _w (n), and _hh (・) is the autocorrelation sequence of h _w (n). It is expressed as follows. Note that _hh (•) is also called the covariance function.

The conventional method determines the amplitude and position of the k-th pulse by considering g _{k in} (8) as a function of l _K only.
That is, l _k that maximizes | gk | in (8) is the position of the kth pulse, and g _k at that time is the amplitude of the kth pulse. In this method, if g _k is a function of exactly l _k, the sound source pulse sequence that minimizes Eq. (7) is calculated, but this is not the case for the actual speech signal, and in general g _k is
Functions such as l ₁ , l ₂ , ..., l _k .

FIG. 1 is a block diagram showing the conventional method of Document 2.
FIG. 2 is a flowchart showing a processing procedure for obtaining the amplitude g _k and the position l _k of the sound source pulse sequence performed by the sound source pulse sequence calculation circuit 140 according to the conventional method of Document 2. Hereinafter, the constituent elements of the embodiment of the literature 2 conventional system shown in FIG. 1 and the source pulse sequence search algorithm by the literature 2 conventional system shown in FIG. 2 will be described in detail. In FIG. 1, each constituent element performs processing for each frame. 100 indicates an encoder input terminal, A /
The D-converted audio signal sequence x (n) is input. Reference numeral 110 denotes a buffer memory circuit, which stores an audio signal sequence for one frame. The K parameter calculation circuit 180 inputs the audio signal x (n) accumulated in the buffer memory circuit 110 and calculates K parameters K _i (1 ≦ i ≦ M) by a predetermined number. This value is output to the K parameter encoding circuit 190. The K parameter encoding circuit 190 encodes K _i based on, for example, a predetermined number of quantization bits, and outputs the code I _ki to the multiplexer 160. Further, the K parameter encoding circuit 190 decodes I _ki and outputs the decoded value K ′ _i.
(1 ≦ i ≦ M) is output to the impulse response calculation circuit 120 and the weighting circuit 200. The weighting circuit 200 inputs the input speech signal x (n) and the K parameter decoded value K ′ _i , and uses the weighting function w (n) depending on the frequency characteristic of the synthesis filter to calculate the above x _w (n). Then, the obtained x _w (n) is output to the cross correlation coefficient calculation circuit 135. The impulse response calculation circuit 120 uses K ′
_i is input, the above-mentioned h _w (n) (convolution of the impulse response and the above-mentioned weighting function) is calculated by a predetermined number of samples, and the obtained h _w (n) is calculated by the covariance function calculation circuit 130. It outputs to the cross-correlation function calculation circuit 135. Covariance function calculation circuit 1
For 30, input h _w (n) of a predetermined number of samples, and according to the above equation (10), _hh (l _i , l _j ) (0 ≦ l _i , l _j ≦
N-1) is calculated, and this is used as the sound source pulse sequence calculation circuit 140
Output to. The cross-correlation function calculation circuit 135 uses the input x
The cross-correlation function between _w (n) and h _w (n) is calculated and output to the sound source pulse sequence calculation circuit 140. Next, the sound source pulse sequence calculation circuit will be described. The sound source pulse sequence calculation circuit 140 uses the cross correlation _coefficient calculation circuit 135 to _obtain ψ _xh (l _k ) (0 ≦ l _k ≦ N−1)
From the covariance function calculation circuit 130 to _hh (l _i , l _j ) (0 ≦ l _i ,
l _j ≦ N−1) is input, and the amplitude g _k and the position l _k of the sound source pulse sequence are calculated using the above-described pulse calculation algorithm (8). FIG. 2 is a flowchart showing a processing procedure performed by the sound source pulse sequence calculation circuit 140 in the conventional method of Document 2. The first pulse is in equation (8),
K = 1 and amplitude g ₁ is a function of position l ₁ , g ₁ = ψ
_It is represented as _xh (l ₁ ) / _hh (l ₁ , l ₁ ). Then | g ₁
Select l ₁ that maximizes |, and let l ₁ and g ₁ at that time be the _first pulse position and amplitude. The second pulse, (8) K = 2 Distant in formula, | g ₂ | Select a l ₂ to maximize, to the position and amplitude of the l _2, g ₂ at that time the second pulse. The third and subsequent pulses are calculated in the same manner, and are continued until the number of pulses determined in advance is reached. In FIG. 2, 1 initializes a calculation counter for calculating the number of pulses to 1. Reference numeral 2 is a comparison, and it is determined whether the number of pulses is larger or smaller than a predetermined number, and if it is larger than the predetermined number, the pulse sequence calculation process is terminated. Three
Is to calculate equation (8), and in equation (8), l ₁ ,…,
l _k-1 and g _1, ..., the g _k-1 and known, | g _k | a to maximize l _k
And g _k , l _k at that time are output as the amplitude and position of the k-th pulse. Reference numeral 4 is an adder, which increases the content of a calculation counter for calculating the number of pulses. This completes the description of the sound source pulse calculation circuit 140.

Returning to FIG. 1, the encoding circuit 150 inputs the amplitude g _k and the position l _k of the pulse sequence which is the output of the excitation pulse calculation circuit 140, and encodes them. A method well known in the related art can be used for encoding the amplitude g _k and the position l _k . Regarding the amplitude g _k , for example, a method in which the maximum value of the amplitude of the pulse sequence in one frame is used as a normalization coefficient and the amplitude of each pulse is normalized with this value, and then quantization and encoding are considered. For position l _k , it is conceivable to use run-length coding, which is well known in the field of facsimile signal coding, for example. This represents the length following the code “0” using a predetermined code sequence. The multiplexer 160 inputs the output code of the K parameter encoding circuit 190 and the output code of the encoding circuit 150,
These are combined and output from the transmission-side output terminal 170 to the communication path.

(Problems of Prior Art) The drive source pulse sequence search method proposed in the conventional method of Document 2 has been described above. Document 2 The conventional method is based on the assumption that the pulse amplitude is a function of only the position where the pulse stands in the algorithm for obtaining the amplitude and position of the sound source pulse sequence. However, the above assumption does not hold for the actual voice signal, and g in the above equation (8) used to obtain the sound source pulse sequence in the conventional method in Reference 2 is used.
_k is generally a function such as l ₁ , ..., l _k . Therefore, the sound source pulse sequence determined by the conventional method of Document 2 is
There is a more suitable sound source pulse sequence that does not make J in equation (7) really small. In the conventional method, the amplitude of the sound source pulse sequence is all determined and then quantized. With such quantization, it is not possible to scoop out a quantization error caused by amplitude quantization. Further, in such a method of directly quantizing the source pulse sequence, the quantization characteristic greatly depends on the quantization width of the amplitude of the source pulse sequence, and in order to obtain good quantization characteristic, the amplitude of the source pulse sequence is Many bits have to be allocated. Therefore,
In order to obtain better voice quality in a region where the transmission rate is about 16 kbit / sec or less in the method of expressing the driving sound source signal sequence by a plurality of pulses, it is necessary to find a more suitable amplitude and position of the sound source pulse sequence. , It is necessary to use the source pulse search algorithm that scoops the quantization noise generated by the amplitude quantization.

(Object of the Invention) It is an object of the present invention to provide a high-quality speech coding system suitable for a transmission rate of about 16 Kbit / sec.

(Structure of the Invention) According to the present invention, a discrete audio signal sequence is input, a short-time audio signal sequence obtained by dividing the audio signal sequence for each short time is obtained, and a parameter representing a spectrum envelope from the short-time audio signal sequence is obtained. Is extracted and encoded, an impulse response sequence corresponding to the spectral envelope is calculated, and when a parameter describing a sound source pulse sequence suitable as a driving sound source signal sequence of the short-time speech signal sequence is sequentially obtained, The position of the sound source pulse newly determined by using the short-time voice signal sequence while sequentially orthogonalizing the impulse response sequence having a phase delay corresponding to the position of the sound source pulse defined in, and the orthogonalized signal sequence A parameter that describes the source pulse sequence by calculating and quantizing a product sum over a predetermined time with the short-time signal sequence. The position of the excitation pulse and the quantized sum of products are obtained, the excitation pulse and the quantized sum of products are encoded, and the code of the parameter representing the spectral envelope and the excitation pulse sequence are described. A speech coding method characterized in that the discrete speech signal sequence is coded by combining with the code of the parameter.

Further, according to the present invention, a discrete voice signal sequence is input, a short-time voice signal sequence obtained by dividing the voice signal sequence for each short time is obtained, and a parameter representing a spectrum envelope is extracted from the short-time voice signal sequence. Calculate an impulse response sequence having a spectrum that is encoded and has a predetermined correction added to the spectrum envelope, and calculates a short-time speech signal sequence in which the predetermined correction is added to the short-time speech signal sequence. While sequentially obtaining a parameter describing a sound source pulse sequence suitable as a driving sound source of the short-time speech signal sequence, sequentially orthogonalizing an impulse response sequence with a phase delay corresponding to the position of a newly determined sound source pulse. The position of the sound source pulse newly determined using the corrected short-time audio signal sequence is determined, and the corrected short-time audio signal system is added. And a product sum of the orthogonalized signal sequence over a predetermined time is calculated, the product sum is quantized, the quantized product sum and the position of the determined excitation pulse are encoded, and the spectrum is obtained. A speech coding method, wherein the discrete speech signal sequence is coded by combining a code of a parameter indicating an envelope, a code indicating a position of the excitation pulse, and a code indicating the quantized sum of products. Is obtained.

(Principle of the Invention) The speech coding method according to the present invention is characterized by the method of expressing the excitation pulse sequence, the algorithm for obtaining them, and the quantization method. Therefore, when the following equation (7) is given, the amplitude g _k , k = of the source pulse sequence that minimizes J is given.
, ..., K and the positions l _k , k = 1, ..., K are sequentially determined by the algorithm of the present invention.

Expression representing weighted squared error when K pulses are added Is partially differentiated by g _k , k = 1, ... Where the inner product and the squared error are (12) is expressed as Substituting the relationship of Eq. (15) into Eq. (11), Becomes In equation (11), h _w (nl _k ), k = 1, with different phases
…, The group of K {h _w (nl _k )} generally does not form an orthogonal system. That is, <h _w (nl _i ), h _w (nl _j )> ≠ 0 and i ≠ j− (17). Therefore, reduce J in Eq. (11) {l
_Consider sequentially converting {h _w (nl _k )} into an orthogonal sequence {η _k (n)} in order to sequentially find _k } with respect to k. The use of Schmitt (SCHIMDT) orthogonalization for this successive transformation is as follows.

This Schmidt orthogonalization is from h _w (nl _k ) to {h _w (nl _i )}, i =
Equivalent to removing the correlation with 1, ..., k-1. {Η
_{Since k} (n)} has the following orthogonal relationship <η _i (n), η _j (n)> = 0 i ≠ j − (19), {η _k (n)} holds x _w (n) ) Is a linear least squares approximation, the error is (Written by Shin Ichimatsu, Kikishiki, page 24, Takeuchi Shoten (Sho 3
8), reference 3). Here, if we further set ξ _k = <x _w (n), η _k (n)> − (21), equation (20) becomes Is expressed as

If it is assumed that l ₁ , ..., L _k-1 are determined in the sequential process, it will be calculated up to η ₁ (n), ... η _k-1 (n) from the recurrence formula of Eq. (18). Become. Therefore, the k-th pulse position l _k
Is to minimize the squared error in Eq. (22), that is, Is determined to be the maximum.

In addition, ξ _k is calculated from Eqs. (18) and (21) Therefore, equation (23) is Is equivalent to Therefore, (21) and (23) and the xi] _k quantizes each time the xi] _k, l _k is determined from, seek l _k (2
If quantized ξ _i , i = 1, ..., K-1 is used in equation (5), a position l _k that takes into account the quantization effect of ξ _i , i = 1, ..., k-1 can be obtained. . Quantized ξ _k In other words, from equation (25), l _k is _obtained by maximizing the following equation.

The present invention was obtained as described above. , And a code representing l _k , k = 1, ..., K are used as transmission parameters.

On the other hand, if ξ _k , k = 1, ..., K and l _k , k = 1, ..., K are determined, g _k , k = 1, ..., K are calculated as follows. First, from the comparison between Eqs. (16) and (20) There is a relationship. In this equation, {h _w (nl _k )} in equation (18)
Between {η _k (n)} and Substituting However, b _ii = 1 and b _ij = 0 i <j. By comparing both sides of equation (29), Therefore, using {ξ _k } gives {g _k } Calculated by At the receiving side, the encoded ξ _k , k
, K and l _k , k = 1, ..., K are received, and they are decoded to calculate g _k , k = 1, ..., K from the equation (31). This is the end of the description of the algorithm of the present invention.

(Embodiment) An embodiment of the voice encoding system according to the present invention will be described with reference to the drawings. FIG. 3 (a) is a block diagram of the transmitting side, and FIG. 3 (b) is a block diagram of the receiving side. In FIG. 3 (a), reference numeral 500 denotes an encoder input terminal to which a discrete audio signal sequence x (n) is input. A buffer memory circuit 310 stores one frame of the audio signal sequence. 320 is a K parameter calculation circuit, which is the audio signal x (n) accumulated in the buffer memory circuit 320.
, And calculate K parameters by a predetermined number. This value is output to the K parameter encoding circuit 330. The K parameter encoding circuit encodes the K parameter based on a predetermined number of quantization bits and outputs it to the multiplexer 380. Further, the K parameter coding circuit decodes the coded K parameter and outputs the decoded value to the weighting circuit 340 and the impulse response sequence calculation circuit 350. The weighting circuit 340 inputs the input speech signal x (n) and the decoded value of the K parameter from 330, and weights the weighting function w (n) depending on the frequency characteristic of the synthesis filter.
X _w (n) (voice signal sequence x (n) and weighting function w
(convolution with (n)) is calculated and output to the sound source pulse sequence parameter calculation circuit 360. The impulse response sequence calculation circuit 350 inputs the decoded value of the K parameter from 330 and inputs the above-mentioned h _w (n) (convolution of the impulse response sequence h (n) of the synthesis filter and the weighting function w (n)). Is calculated for a predetermined number of samples, and the obtained h _w (n) is output to the source pulse sequence parameter calculation circuit 360. Next, the sound source pulse sequence parameter calculation circuit 360 will be described. This circuit, x _w (n) is from the weighted impulse response series calculating circuit 350 h _w (n) is inputted from the weighted circuit 340, the algorithm described above (18), (21), (25) The parameters {l _k }, {ξ _k } representing the sound source pulse sequence are calculated using the formula. FIG. 4 shows a parameter calculation circuit 36 for the sound source pulse sequence.
6 is a flowchart showing a processing procedure performed in 0. Reference numeral 5 is for setting an initial value, and is for calculating a value with k = 1 in the equations (18), (21) and (26). (18) eta _{1 (n)} and (21) from _{ξ 1 = <x w (n} ), η 1 (n)> from the equation to calculate the (26) from equation _{^{▲ ξ 2 1 ▼ / <η}} 1 Determine l ₁ that maximizes (n), η ₁ (n)>. Reference numeral 6 denotes addition, which adds one value of k representing the number of pulses. Reference numeral 7 is a comparison, and it is determined whether the calculated pulse number is larger or smaller than a predetermined number, and when it is larger than the predetermined number, the process of calculating the pulse position is stopped. 8 is the above formula (18) and (26)
Equation (18) is used to calculate η _k (n), and Equation (26) is used to calculate To calculate. 9 is for obtaining the position l _k of the sound source pulse,
Let l _k which maximizes the above equation (26) be the position of the sound source pulse.
10 xi] _k and requests, the (21) get xi] _k quantizes it computes the xi] _k from the equation. Various methods are conceivable for the quantization of {ξ _k }. For example, the first obtained | ξ
₁ | is used as a normalization coefficient and ξ ₁ is obtained from the following {ξ _k } is normalized and sequentially uniformly quantized, or | ξ ₁ | is used as an initial value for | ξ _i-1 | and | ξ _i | i = A method in which the difference between 2, ..., K is sequentially quantized and the code is stored can be considered.
This completes the description of the source pulse sequence parameter calculation circuit 360.

Returning to FIG. 3 (a), the encoding circuit 370 is the output of the excitation pulse sequence parameter calculation circuit.

Input and encode them.

There are various conceivable ideas for the encoding of. But, Since is a value determined by successive orthogonal transformation, it is necessary for the decoding side to know the order of orthogonalization at the time of encoding. As an example, A method of run-length encoding {l _k } in the order corresponding to As another example, If i <j, then Since there is an order relation such that {l _k } is rearranged in an order that is easy to encode, On the decoding side According to equation (32) An encoding method is considered in which {l _k } is returned to the original order by transforming in the order of magnitude. However, Because there is a possibility that A code must also be assigned to the code of. The multiplexer 380 inputs the output code of the K parameter encoding circuit and the output code of the encoding circuit 370, combines them, and outputs them from the transmission side output terminal 510 to the communication path.

Next, the receiving side shown in FIG. 3 (b) will be described. The demultiplexer 390 inputs a code through the reception side input terminal 520,
The code representing the K parameter and the code representing the excitation pulse sequence are separated, and the code representing the K parameter is decoded by the decoder 40.
The code representing the excitation pulse sequence is output to 0 to the decoder 410. The decoder 400 decodes the code representing the K parameter input from the demultiplexer 390, and outputs it to the impulse response sequence calculation circuit 420 and the audio reproduction circuit 450.
Decoder 410 receives the code representing the excitation pulse sequence from demultiplexer 390, and outputs the parameters of the excitation pulse sequence.
Decode into {l _k } and {ξ _k }. The impulse response sequence calculation circuit 420 inputs the decoded K parameter, calculates the above-mentioned weighted impulse response sequence h _w (n), and outputs it to the orthogonal transformation circuit 430. The orthogonal favorable circuit 430 inputs the weighted impulse response sequence h _w (n) and the output {l _k } of the decoder 410 and inputs the orthogonal sequence {η by the recurrence formula of the equation (18).
_k (n)} and the transformation matrix {b _ij } shown in the equation (28) are calculated. The sound source pulse amplitude calculation circuit 440 is the orthogonal transformation circuit 430.
From {η _k (n)}, {b _ij }, which is the output of { _{circumflex over} ()}, and {ξ _k } that is the output of the decoder 410, the amplitude {g _k } of the sound source pulse is calculated using the above equation (29). , And outputs it to the audio reproduction circuit 450. The audio reproduction circuit 450 calculates a synthesis filter from the K parameter that is the output of the decoder 400, and outputs the driving excitation sequence that is the input of the synthesis filter to the output {l _k } of the decoder 410 and the output of the excitation pulse amplitude calculation circuit { g _k }, and the calculated driving sound source sequence is added as an input to the calculated synthesis filter to calculate a re-voiced voice signal sequence and output to the output terminal 530.

The embodiment of the present invention has been described above. In the example described here, <x _w (n), η _k (n)> is used as {ξ _k }, but <x _w (n), η _k (n)> is included as ξ _k. Anything can be used, for example, <x _w (n), η _k (n)> / | η _k
(n) | or <x _w (n), η _k (n)> / <η _k (n), η _k (n)>. Also, the impulse response sequence of the synthesis filter
h _w (n) decays exponentially. Therefore, it can be said that the correlation between h _w (nl _i ) and h _w (nl _j ) is small where | l _i -l _j | is large. Therefore, in the recurrence formula (18), | l _i -l _j |
If the value of is larger than the predetermined value, (18)
An approximate orthogonalized sequence {η _k (n)} can be calculated without performing the correlation removal operation of the equation. Such a configuration can significantly reduce the calculation required for orthogonalization.

Further, in the algorithm described in the action-the principles of the present _{invention, <x w (n),} h w (nl i)> = ψ xh (l i), <h w (n-
Since l _i ), h _w (n−l _j )> = _hh (l _i , l _j ), even if ψ _xh (l _i ), _hh (l _i , l _j ) is calculated first, The invention can be realized. The {b _ij }, the {<< η _k (n),
η _k (n)>, the above {ξ _k }, the above {l _k } and ψ _xh (l _i ),
The relation with _hh (l _i , l _j ) can be expressed as follows. First,
From the equation (18) and the equation (19) Next, from equation (26) above By using the equation (31) and the equation (32), ψ _xh (l _i ) and _hh
From (l _i , l _j ), {l _k }, which is an important parameter in the present invention,
The algorithm for obtaining {ξ _k } and {g _k } is patent application number Sho
58-150783 "Voice coding method" (reference 4). {V _ij } in the equations (17) and (18) in the reference 4 is equal to {b _ij } in the equation (18) in this specification. (19) in Reference 4
In the present specification, {d _k } of the equation (20) is {<< η _k (n), η
is equal to _k (n)>}. Also, Equations (25) and (26) in Reference 4
The expression {y _k } is equal to {ξ _k } in the expression (21) in the present specification. Further, the equation (28) in the reference 4 for obtaining the position l _k is equal to the equation (23) in this specification. Further, in order to obtain the amplitude {g _k }, the equations (31) and (32) in Reference 4 are equal to the above equation (31) in this specification.

The calculation of the sound source pulse sequence of the present invention has been performed frame by frame, but the frame may be divided into several subframes and the pulse sequence may be calculated for each subframe. According to this configuration, the amount of calculation can be increased by about 1 / m as compared with the configuration shown in FIG. 3 in which the number of frame divisions is m.

Further, although the frame length is fixed in the configuration example described above, it may be variable. The characteristics can be improved by making it variable. The K parameter is used as a parameter representing the spectrum envelope of the short-time speech signal sequence, but this is another well-known parameter (for example, L parameter).
SP parameters, etc.) may be used. Furthermore, the weight function w (n) described above is not an element necessary for implementing the present invention, and may be omitted. However, the effect of the present invention can be further enhanced by adding the weighting function w (n) in consideration of human auditory characteristics.

(Effect of the Invention) According to the configuration of the present invention, l ₁ , ..., L _k-1 and ξ ₁ , ...,
In order to sequentially find the optimum position l _k considering the quantization of ξ _k−1 , and to determine the optimum amplitude considering the quantization of the positions {l _k } and {ξ _k } by equation (31), Unlike observing the amplitude of the pulse as a function only in the position where the pulse stands as in the conventional method of Document 2, a more suitable sound source pulse sequence can be obtained in the sense of reducing the square error. Therefore, there is an effect that better sound quality can be obtained as compared with the conventional method. Further, as in the configuration of the present invention, the one including the quantization of the amplitude of the orthogonal sequence in the sequential process of sequentially orthogonalizing the impulse response sequence, instead of quantizing the amplitude of the impulse response sequence, There is an effect that the quantization error caused by the quantization can be compensated for in the sequential quantization process, and the quantization characteristic is superior to the conventional method.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment for realizing a conventional method, FIG. 2 is a flow chart showing a processing procedure performed by a sound source pulse sequence calculation circuit according to the conventional method, and FIGS. 3 (a) and 3 (b) are the present invention. FIG. 4 is a block diagram showing an embodiment of the present invention, and FIG. 4 is a flow chart showing a processing procedure performed by a sound source pulse sequence parameter calculation circuit in the present invention. In the figure, 110,310 ... buffer memory circuit, 120,35
0,420 …… Impulse response sequence calculation circuit, 130 …… Covariance function calculation circuit, 135 …… Cross-correlation sequence calculation circuit, 140…
… Source pulse sequence calculation circuit, 150,370 …… Coding circuit, 1
60,380 …… Multiplexer, 180,320 …… K parameter calculation circuit, 190,330 …… K parameter encoding circuit, 200,3
40 …… Weighting circuit, 360 …… Sound source pulse sequence parameter calculation circuit, 390 …… Demultiplexer, 400,410 ……
Decoder, 430 ... Orthogonal transformation circuit, 440 ... Sound source pulse amplitude calculation circuit, 450 ... Voice reproduction circuit, 1 ... Initialization, 2 ...
… Comparison 3 …… Pulse calculation, 4 …… Addition, 5 …… Initialization, 6 …… Addition, 7 …… Comparison, 8 …… Parameter calculation of the sound source pulse sequence by successive orthogonalization, 9 …… Maximum value detection ,
10 ... Shows parameter calculation and quantization of a sound source pulse sequence, respectively.

Claims (2)

[Claims] 1. A short-term audio signal sequence obtained by inputting a discrete audio signal sequence and dividing the audio signal sequence for each short time,
A parameter representing a spectrum envelope is extracted and encoded from the short-time speech signal sequence, an impulse response sequence corresponding to the spectrum envelope is calculated, and a sound source pulse sequence suitable as a driving sound source signal sequence of the short-time speech signal sequence is obtained. When sequentially determining the parameters to be described, the position of the sound source pulse newly determined by using the short time voice signal sequence while sequentially orthogonalizing the impulse response sequence having a phase delay corresponding to the position of the newly determined sound source pulse Decide
The position and the quantum of the sound source pulse which is a parameter describing the sound source pulse sequence by calculating and quantizing a product sum of the orthogonalized signal sequence and the short time signal sequence over a predetermined time. The sum of the product sums, the excitation pulse and the quantized sum of products are encoded, and by combining the sign of the parameter representing the spectral envelope and the sign of the parameter describing the excitation pulse sequence, A speech coding method characterized by coding a discrete speech signal sequence. 2. A discrete-time audio signal sequence is input, and the audio signal sequence is divided into short-time audio signal sequences to obtain a short-time audio signal sequence,
A parameter representing a spectrum envelope is extracted and encoded from the short-time speech signal sequence, an impulse response sequence having a spectrum in which a predetermined correction is added to the spectrum envelope is calculated, and the short-time speech signal sequence is preliminarily described. The position of the newly determined sound source pulse is calculated when a short-time sound signal sequence to which a predetermined correction is added is calculated and the parameters describing the sound source pulse sequence suitable as the driving sound source of the short-time sound signal sequence are sequentially obtained. The position of the sound source pulse newly determined by using the short-time speech signal sequence to which the correction has been added while sequentially orthogonalizing the impulse response sequence having a phase corresponding to Compute a product sum over a predetermined time with the orthogonalized signal sequence, quantize the product sum, and quantize Encoding the sum of products and the position of the determined excitation pulse, the discrete by combining the code of the parameter representing the spectral envelope, the code indicating the position of the excitation pulse and the code indicating the quantized sum of products A speech encoding method characterized by encoding a dynamic speech signal sequence.