JPS62207036A - Voice coding system and its apparatus - Google Patents

Abstract

PURPOSE:To synthesize voice with high quality even with comparatively less operation quantity and low transmission bit rate by using a spectrum parameter obtained as to a representative period and a pulse train so as to represent the voice signal of the entire frame in an excellent way. CONSTITUTION:A drive signal restoration circuit 240 uses a decoded value, position of a representative period and an interpolated pitch of an inputted sound source pulse train, recovers the pulse train of the sub-frame period other than the representative period from the preceding and succeeding frame pulse trains by the interpolation processing to generate a sound source signal for one frame, which is outputted to a synthesized filter circuit 250 as a drive sound source signal. The circuit 250 inputs a drive sound source signal and a K parameter interpolated based on the representative period, converts the K parameter into a forecast coefficient while switching the parameter at each sub-frame and calculates a reply signal. A multiplexer circuit 260 inputs a code k1 representing the K parameter of the representative period, a code ld of a pitch coding circuit 150, a code of a coder 230, a code representing the sub frame phase and a code representing the position of the representative period and combines them to output the result from a terminal 270.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to an audio encoding method and an apparatus thereof, and in particular to an encoding method and its method for encoding an audio signal with low and high quality traits. Regarding equipment.

[Conventional technology]

If the audio signal is transmitted at a low transmission bit rate (e.g. 4°8K)
A vocoder (approximately
VOCODER) is known. The principle of this method is described, for example, by M.R. Schreiter (M.R.
, 5CHROBDER) by °゛Vocodaz: Analysis and Synthesis of Speech °' (
“VOCODER3: ANALYSIS AND
5YNTHESISOF 5PEECH”) (PROC, IEEE, p, p, 720-734.
MAY.

1966) (Reference 1).

In addition, the LPC VOCODER is known as a vocoder that uses the linear predictive analysis method.
Linear Prediction Vocoder Beist Avon The Autocorrelation Method "'("AL
INEARPREDI(':1'lON VOCOD
ERBASED UPON THE AUTOC
A paper entitled "ORRELATION METHOD") (IEEE TRANS, A, S, S, P,.

ρ, p, 124-134. APRIL, 1
974) (Reference 2). The present invention is an improvement of the sound source section of the VOCODER and is closely related to the LPCVOCODER, so an outline of the LPCVOCODER will be explained below, focusing on the configuration of the synthesis section.

FIG. 3 is a block diagram of the combining section (receiving section) of the LPC VOCODER described in Document 2, which shows an example of the combining side of the conventional system. The synthesis section consists of a sound source generation section 500 and a synthesis filter circuit 510. The sound source generator 500 includes an impulse generator 501, a noise generator 502, and a voiced/unvoiced switching circuit 5.
03 and a gain circuit 504. VOCOD
In ER, audio signals are divided into two types, voiced and unvoiced, at short intervals (for example, 20 m5ec), and in the case of voiced signals, the impulse generator 501 generates a pulse train with a time interval of the Sochi period Pd. On the other hand, if there is no voice, white noise is generated from the noise generator 502. Voiced/unvoiced control is performed by a switching circuit 503. A gain G is applied to the signal generated in this manner by a gain circuit 504, and the signal is outputted to a synthesis filter circuit 510 as a sound source signal d(n).

The synthesis filter circuit 510 synthesizes and outputs the sound x (n>) using the sound source signal d (n) and the filter parameters. Here, the pitch period Pd, voiced/unvoiced switching signal (V/1JV), gain G , + to the filter parameter
is calculated at predetermined time intervals on the analysis side (sending side) and transmitted to the receiving side.

In the LPC VOCODER described above, the transmission information is the pitch period, voiced/unvoiced switching signal, gain, and filter parameters, and since the audio signal can be synthesized from these information, the transmission bit rate can be set to a low value (for example, 4.8
kbps). However, with this conventional method, it is difficult to synthesize a good quality voice.

The reason is that when the sound source signal is voiced, the sound source is represented by one impulse per bi-sochi, and it also does not include phase information, so the naturalness is considerably impaired, and the synthesized sound is so-called mechanical sound. It was a sound.

In addition, since the sound is divided into two extremes, voiced and unvoiced, and the sound source is switched between an impulse sound source and a noise source, there is a drawback that if an error in voiced/unvoiced discrimination occurs, it will cause a large quality deterioration. there were. Furthermore, at the transition between voiceless and voiced, the sound source could not be expressed well, resulting in deterioration. Furthermore, if the pitch period is shifted and cannot be determined, there is a drawback that a large quality deterioration is caused.

As a method of improving the sound source, for example, Japanese Patent Application No. 59-272
As described in Specification No. 435 (Document 3) etc.,
A method is known in which a pulse train of one pitch section is used to represent a sound source for an entire frame. In this method, on the transmitting side, one frame is divided into pitch sections for each pitch period using the pitch period found in the frame, and a spectral parameter (combined A pulse train is obtained for each pitch section using the filter coefficients, and one pitch section (representative section) that can reproduce a good signal for the entire frame is selected.

[Problem that the invention seeks to solve]

In the method of Reference 3 mentioned above, the spectrum parameter is determined to represent the average characteristics for the entire frame, regardless of the position of the representative section, and in this case, it is possible to use a frame time length of about 20 m5 eC. many,
In frames where the spectral characteristics of the voice change greatly over a short period of time, such as in transient parts or vowel transition parts, the spectral parameters also change over a short time. There was a problem in that the spectral parameters determined from the characteristics were not necessarily optimal in the representative section.

SUMMARY OF THE INVENTION An object of the present invention is to provide a speech encoding method and apparatus that can synthesize high-quality speech even at a low transmission bit rate with a relatively small amount of calculation.

[Means for solving problems]

In the audio encoding method of the first invention, on the transmitting side, a discrete audio signal is input, divided into predetermined first time intervals, and a pitch parameter representing the pitch is extracted from the audio signal. dividing the time interval into second time intervals shorter than the time interval, selecting one interval from the second time interval, and selecting a spectral parameter representing the short-time spectral envelope of the audio signal of the selected interval; Information representing the sound source of the audio signal is obtained, and the spectral parameter and sound source information are combined and output with the pitch parameter, and the receiving side restores the second time interval based on the pitch parameter and the selected sound source. The audio signal is synthesized using information representing the sound source of the section and spectral parameters of the selected section.

Further, the speech encoding device of the second invention divides an input speech signal into a predetermined first time interval, extracts and encodes a pitch parameter representing a pitch from the speech signal, and also extracts and encodes the pitch parameter. and dividing the audio signal into a second time interval shorter than the first time interval, and selecting one interval of the second time interval and dividing the audio signal of the selected interval into a short time interval. a parameter calculation circuit that calculates and encodes a spectral parameter representing a spectral envelope and a pulse train representing a sound source of the audio signal in the selected section; a code representing the sound source of the selected section and a spectrum of the selected section; It has a multiplexer circuit that outputs a combination of a code representing a parameter and a code representing the pitch parameter.

Furthermore, the audio decoding device of the third invention includes a code representing a pitch parameter for each predetermined first time interval, a code representing a spectral parameter of the selected interval, and a code representing a sound source of the selected interval. a demultiplexer circuit which inputs a code sequence in which the above is combined, and separates and decodes a code representing the pitch parameter, a code representing the spectral parameter of the selected section, and a code representing sound source information of the selected section; a dividing circuit that restores a second time interval shorter than the first time interval based on the decoded pitch parameter; a driving sound source signal restoration circuit that restores the driving sound source signal by applying processing to give a smooth change to the driving sound source signal; and a synthesis filter circuit for output.

[Effect]

The present invention is characterized in that the audio signal of the entire frame is well represented using the spectral parameters and pulse train determined for one pitch section (representative section).

To implement this method, for example, find the spectrum parameters for each pitch section, use this to calculate a balsa sequence for each Bi-Sochi section, and search for sections that can reproduce a good audio signal over the entire frame. , a method of transmitting the pulse train of the representative section and the spectral parameters of the representative section is conceivable. Another possible method is to select a representative section using an appropriate method and then obtain the spectral parameters and pulse train of this section. FIG. 2 shows an example of processing of the former method. <a) shows the audio waveform of one frame, and (b) shows the frame divided into pitch sections. Here, spectrum parameters are determined for each pitch section. (C) shows a pulse train obtained for each pitch section using this spectrum parameter, and (d) shows a selected representative section and a pulse train in that section.

Here, as a method for selecting the representative section, the same method as shown in the above-mentioned document 3 can be used.

Further, as a method for determining the amplitude and position of the pulse train in the pitch section, the method described in the above-mentioned document 3 can be used, but in addition to this, for example, analysis by synthesis (ANALYSIS I 5-by-3YNTHES I
A method is known that uses the method of S;
A new model of
LBC Excitement for Producing Natural Sounding Speech at Low Bit Rates'("ANEWMOD")
EL OF LPC EXCITATION FO
RPRODtJCING NATURAL 5OU
NDING 5PEECHATLOW BIT
RATES”) (PROC, 1, C, A,
S, S, P,, P, I).

614-617.1982> (Reference 4).

〔Example〕

Next, the present invention will be explained in detail with reference to the drawings.

FIGS. 1(a) and 1(b) are block diagrams of a transmitting side and a receiving side, respectively, in an embodiment of the audio encoding method of the present invention.

In FIG. 1(a), an audio signal x (n>) is input from the transmission side input terminal 100, and a frame (first time interval)
Each time, a predetermined number of samples are stored in the buffer memory circuit 110.

Next, the parameter calculation circuit 140 inputs a predetermined number of samples of the audio signal from the buffer memory circuit 110 and calculates parameters representing the average spectral envelope of the audio signal within the frame. The parameters here are the same as the PARCOR coefficients. K
The autocorrelation method is well known as a parameter calculation method. For more information on this method, please refer to John Makhoul et al.'s Ontization Properties of Transmission Parameters in Linear Prediction
Systems (“'QUANTIZATION PRO
PERTIBS OF TRANSMISSION
PARAMETER3INLINEARPREDI
CTIVE SYSTEMS”) (IE
EE TRANS,A,S,S,P,,ρ,p,30
9-321° 1983) (Reference 5), so the explanation will be omitted here. Returning to FIG. 1(a), the K parameter is output to the parameter encoding circuit 160.

The K parameter encoding circuit 160 encodes + into the parameter based on a predetermined number of quantization bits, further decodes it, converts the obtained parameter decoded value 1° into a prediction coefficient a1°, and converts it into a subframe dividing circuit. Output to 165.

The pitch analysis circuit 130 uses the output of the buffer memory circuit 110 to calculate the pitch period Pd within the frame. The method for calculating the pitch period Pd is, for example, the ``Real Time Imbrimention of Time Domain'' by R. V. Cox (R, V, C0X) et al.
Harmonics Scaling of Speech ("RE")
AL-TIME IMPLEMENTATION
OFTTMF, DOMAIN HARMONIC
3CALING OF 5PEECH5IGNAL
A paper entitled “S”) (IEEE TRANS.

A, S, S, P,, p, p, 258-272.1983
> (Reference 6) etc. can be used.

The Pisochi encoding circuit 150 quantizes and encodes the pitch period Pd using a predetermined number of quantization bits, and
d to multiplexer 260. Furthermore, the pitch period Pd' obtained by decoding is output to the pitch interpolation circuit 225.

The pit interpolation circuit 225 interpolates the pitch period using the pitch period Pd' and the pitch period of the adjacent frame, and outputs the interpolated pitch period to the subframe division circuit 165.

The subframe division circuit 165 divides the frame into subframes (second time intervals) for each pitch period using the prediction coefficients a, ゛ and the interpolated pitch period. As a subframe division method, a method is known in which a sound source pulse train representing a sound source signal is obtained and divided, and this method is described as the drive signal calculation circuit 220 in the embodiment of the above-mentioned document 3, so it will not be explained here. omitted. - As an example, the second
Figures (a) and (b) show the audio waveform of one frame and the pattern in which the frame is divided into subframes.9 Here, the subframe phase TP' is expressed by an appropriate bi-soto number code and sent to the multiplexer 260. Output. Further, the subframe division position is outputted to I (parameter calculation circuit 1409, weighting circuit 200, parameter interpolation circuit 2551, synthesis filter circuit 250, drive signal restoration circuit 240, drive signal calculation circuit 220).

Next, the K parameter calculation circuit 140 calculates parameters as spectral parameters for each subframe section according to the input subframe division position, and outputs them to the K parameter encoding circuit 160. The parameters obtained in each case are encoded and decoded, and further converted into prediction coefficients al', which are sent to the impulse response calculation circuit 1701 and the weighting circuit 200. is output to the parameter interpolation port 255. The K parameter interpolation circuit 255 performs linear interpolation between the parameters of adjacent frames using each subframe section as a reference, and obtains interpolated parameters for each subframe.

The impulse response calculation circuit 170 receives as many prediction coefficients a1' as the number of subframes, and calculates impulse responses h w (n) representing the weighted transfer function of the synthesis filter for the number of subframes.

Here, to calculate the impulse response hw(n), for example, the same method as the impulse response calculation circuit 210 described in FIG. 4(a) of Japanese Patent Application No. 59-042305 (Document 7) may be used. I can do it.

The impulse response hw(n) is the autocorrelation function calculation circuit 18
0 and is output to the cross-correlation function calculation circuit 210.

The autocorrelation function calculation circuit 180 inputs the impulse response hw(n) obtained for each subframe, calculates the autocorrelation function coefficient Rhh (m) for each subframe, and outputs it to the drive signal calculation circuit 220. Here, to calculate Rhh (m), for example, the autocorrelation function calculation circuit 1 described in the above-mentioned document 7 is used.
The same method as 80 can be used.

Next, the subtracter 120 subtracts the output of the synthesis filter circuit 250 by one frame from the audio signal X (n) of the buffer memory circuit 110, and uses the subtraction result e (n) as
Output to 00. The weighting circuit 200 calculates the subtraction result e (
n>, divides it into subframes using the subframe division position, and weights the subtraction result e (n) for each subframe using the prediction coefficient a1° for each subframe to obtain a weighted result e w (n) Output 'g. Here, the weighting result ew(n> can be calculated using the same method as the weighting circuit 410 described in FIG. 4 (a>) of Document 7, for example.

The cross-correlation function calculation circuit 210 calculates the weighting result e w (n >) and the impulse response h w (n) for each subframe.
is input to calculate the cross-correlation function φ, and output it to the drive signal calculation circuit 220. Here, to calculate the cross-correlation coefficient φha, for example, the cross-correlation function calculation circuit 2 described in the above-mentioned document 7 is used.
The same method as in 10 can be used.

Next, the drive calculation circuit 220 obtains a sound source pulse train and a spectrum parameter (in this case, a parameter) for one interval that satisfactorily represents the audio signal of one frame. How to obtain the sound source signal will be explained below. First, a frame is divided into subframes as shown in FIG. 2(b) using subframe division positions. Then, using the cross-correlation function input for each subframe section, a predetermined l[ii! Calculate the number of pulses. As a typical subframe interval transition method, for example, focusing on a certain subframe (for example, the second interval ■) in Fig. 2(c),
Linear interpolation is performed between the sound source pulse train of this subframe and the sound source pulse train of an adjacent frame to reproduce the sound source pulse train of other subframes.

Then, the parameters obtained by interpolation using the second subframe section ■ as a reference are inputted from the parameter interpolation circuit 255, and the audio signal of the entire one frame is reproduced using the sound source pulse train and the parameters. Find the error power with the audio signal. The above processing is performed for several subframes, and a subframe section that reduces the error power is selected and used as a representative section. FIG. 2(d) shows the representative section obtained by such a procedure and the sound source pulse train of that section. The amplitude 1 position of the sound source pulse train in the representative section is encoder 2
30. Also, the subframe number of the representative section■
is encoded with a predetermined number of bits and output to multiplexer 260 and parameter encoding circuit 160.

The K parameter encoding circuit 160 inputs the subframe number of the representative section and outputs a code representing the parameter of this section to the multiplexer 260.

Encoder 230 encodes nine amplitude positions of the pulse train and outputs it to multiplexer 260 . Further, the decoded values g,', m,' of the amplitude 1 position of the pulse train are sent to the drive signal restoration circuit 240.
Output to. Here, as the pulse encoding method, for example, the same method as that of the encoding circuit 470 described in Document 7 can be used.

The drive signal restoration circuit 240 uses the position of the decoded value 1 representative section of the input sound source pulse train and the interpolated pitch period to reproduce the pulse train of the subframe section other than the representative section by interpolation processing from the pulse train of the previous and subsequent frames. A sound source signal for one frame is generated and outputted to the synthesis filter circuit 250 as a drive sound source signal.

The synthesis filter circuit 250 inputs the parameters interpolated with reference to the representative section of the driving sound source signal 9, switches the parameters for each subframe, converts them into prediction coefficients, and converts the parameters into prediction coefficients.
A response signal X(n) for one frame is calculated. Here, the same method as in the synthesis filter circuit 400 described in Document 7 can be used to calculate the response signal, for example.

The multi-breaster circuit 260 inputs the code e k l representing the parameter of the representative section, the code /d of the pitch encoding circuit 150, the code of the encoder 230, and the code representing the subframe phase 1 and the code representing the position of the representative section. , these are combined and output from the transmission side output terminal 270. This concludes the explanation of the transmitting side of the audio encoding system of this embodiment.

Next, in FIG. 1(b), the demultiplexer 290
separates the code input from the receiving side input terminal 280 into the code representing the parameter of the representative interval, the code representing the Bi-Sochi period, and the code representing the sound source pulse train, and
Each parameter decoding circuit 330. Pitch decoding circuit 320. It is output to the f3I circuit 300. Further, a code representing the subframe phase is output to the subframe division circuit 355, and a code representing the representative section is output to the drive signal restoration circuit 355.
0 and 1 (output to parameter interpolation circuit 355).

I(parameter decoding circuit 330 decodes the parameter and outputs the decoded value KI° to parameter interpolation circuit 355.

The pitch decoding circuit 320 decodes the pitch period Pd' and outputs it to the pitch interpolation circuit 345. Pisochi interpolation circuit 34
5 determines the interpolated pitch period and outputs it to the subframe division circuit 355.

The subframe division circuit 355 inputs the interpolated pitch period and subframe phase to determine subframe division positions.
Drive signal restoration circuit 340 and parameter interpolation circuit 335
is output to the synthesis filter circuit 350.

The decoding circuit 300 decodes the sound source pulse train and outputs it to the drive signal restoration circuit 340. The drive signal restoration circuit 340 operates in the same way as the drive signal restoration circuit 24o (shown in FIG. 1(a)) on the transmitting side, and generates a sound source pulse train for one entire frame and synthesizes it as a drive sound source signal. Filter circuit 35
Output to 0.

The K parameter interpolation circuit 355 decodes the parameters,
The position of the representative section and the subframe division position are input, parameters are linearly interpolated for each subframe between adjacent frames based on the representative section, and the interpolated parameters are output to the synthesis filter circuit 350. .

The synthesis filter circuit 350 inputs the subframe division position, the driving sound source signal, and the interpolated parameters, and performs the same operation as the synthesis filter circuit 250 on the transmission side (shown in FIG. 1(a)) to generate one frame. Synthetic speech signal ';c (n
) is calculated and output from the receiving side output terminal 360.

This concludes the explanation of the receiving side of the audio encoding system of this embodiment.

In the drive signal calculation circuit 220 of this embodiment, the sound source signal is represented by the sound source pulse train of the representative section, but whether it is voiced or unvoiced is determined for each frame, and in the unvoiced section, the sound source pulse train of the representative section is not used. A sound source may be represented by a combination of a noise source and a pulse train. This is because the sound source signal becomes noisy during unvoiced sections, so this method can improve the sound quality during unvoiced sections. That is, in the case of voiced sound, a pulse train of the representative section may be used as the sound source signal, and in the case of unvoiced sound, a combination of noise and pulse train may be used. This method is described in, for example, the above-mentioned document 3, so the explanation is omitted here.

Here, as a method to easily distinguish between voiced sections and unvoiced sections, as described in the above-mentioned document 3, for example, the pitch gain is calculated from the value of the autocorrelation function 1 pitch apart of the audio signal, and the pitch gain is A method of determining voiced or unvoiced based on the size can be used. Other known methods can also be used.

In addition to the method described in this example, pulse calculation methods include:
Various methods can be used. For example, a method may be used in which the amplitude of previously determined pulses is adjusted each time one pulse is determined. The details of this method are described in the paper titled ``Sound source pulse search method in multipulse-driven speech coding method'' by Mr. Ono et al. (Reference 8) Therefore, the explanation is omitted here.

Additionally, when determining a pulse train, the frame was divided into subframes and then a pulse train was determined for each subframe. It is also possible to use a pulse that falls within a subframe of .

Further, in order to simplify the process, the pulse train may be determined with the parameters constant within a frame. In this case, the K-parameter interpolation circuit 255° 335 becomes unnecessary, but the temporal continuity of the spectral parameters cannot be maintained.

Furthermore, after selecting a representative section in an appropriate manner, for example, near the center of the frame or a section with large power, the pulse train of that section and the parameters of that section may be determined.

Alternatively, it may be divided at predetermined time intervals. In this way, there is no need to transmit the subframe phase.

Further, the position of the representative section may be determined in advance.

In this way, there is no need to transmit the position of the representative section.

Regarding the interpolation of the sound source pulse train and the parameters, interpolation may be performed in synchronization with the pitch period using the selected representative section as a reference, or either one or both of the sound source pulse train and the parameters may be interpolated in a predetermined manner instead of the representative section. Interpolation may be performed on the pitch section (for example, the pitch section near the center of the frame). Alternatively, interpolation may be performed without synchronizing with the pitch period.

Also, the interpolation process for the parameters described in (1) can be omitted, or it can be performed only on the receiving side.

By doing this, the K parameter interpolation circuit 25
5.335 can be omitted.

As the interpolation method for the parameters (spectral parameters) of the sound source pulse train 9 synthesis filter and the pitch period, methods other than linear interpolation may be considered. For example, a sound source array,
Regarding the pitch period, logarithmic interpolation or the like can also be considered. In addition, when interpolating the parameters of the synthesis filter, in this example, the parameters are interpolated. It is also possible to use a method of interpolating parameters or autocorrelation functions. These specific methods are described by Benis Atal (B,
“Speech Analysis and Synthesis by Linear Prediction of the Speech Unit” (“5PEEC”) by Mr. S., ATAL) et al.
HANALYSIS AND 5YNTHESIS
BYLINEAR PREDICTION OF
The paper entitled THE 5PEECHWAVE'") (J, ACOUST, SOC, AM,, p, p.

637-655.1971> (Reference 9), etc., so the explanation will be omitted.

In this example, the parameters were analyzed and the sound source pulse train was calculated assuming that the frame length was constant. However, the frame length may also be variable. Since the frame length can be made shorter and the frame length can be made longer in the stationary part, the transmission bit rate can be reduced. Furthermore, the frame length may be determined according to the pitch period (for example, an integral multiple of the pitch period).

The methods described above may be used alone or in appropriate combinations.

As another configuration method of the present invention, the drive signal restoration circuit 2402 in FIG. 1(a) is combined with the parameter interpolation circuit 255. It is also possible to adopt a configuration in which the subtracter 120 is omitted. In this case, there is no need to synthesize audio signals on the transmitting side, and the device configuration can be simplified.

Note that, as is well known in the field of digital signal processing, the autocorrelation function can also be calculated from the power spectrum. Further, the cross-correlation function can also be calculated from the cross-power spectrum.

Regarding these correspondence relationships, see “Digital Signal Processing” by N.V. Oppenheim et al.
A book titled “PROCESSING” (Reference 10)
Since it is explained in detail in Chapter 8, etc., the explanation will be omitted here.

〔Effect of the invention〕

As described above, the present invention divides one frame into pitch sections, selects one pitch section (representative section), calculates and transmits the spectral parameters and sound source pulse train, and uses these to transmit the audio signal on the receiving side. Since this method uses playback, it has the effect of being able to play back high-quality audio with the same bit rate compared to conventional methods.

[Brief explanation of drawings]

FIGS. 1(a) and (b) are block diagrams of the transmitting side and receiving side, respectively, in an embodiment of the audio encoding method of the present invention, and FIG. 2 is the drive signal calculation circuit shown in FIG. 1(a). FIG. 3 is a block diagram showing an example of the synthesis side of the conventional method. 100... Sending side input terminal, 110... Buffer memory, 120... Subtractor, 130...Pitch analysis circuit, 140...K parameter calculation circuit, 150...
Pitch encoding circuit, 160...K parameter encoding circuit, 165.355...Subframe division circuit, 17
0... Impulse response calculation circuit, 180... Autocorrelation function calculation circuit, 200... Weighting circuit, 210.
... Cross-correlation function calculation circuit, 220 ... Drive signal calculation circuit, 225.345 ... Pitch interpolation circuit, 230.
... Encoder, 240, 340 ... Drive signal restoration circuit,
250°350.510...Synthesis filter circuit, 25
5.335...K parameter interpolation circuit, 260...
Multiplexer, 270... Transmission side output terminal, 280
. . . receiving side input terminal, 290 . . . demultiplexer, 300 . . . decoding circuit, 320 . . . pitch decoding circuit,
330...yffZ diagram (L) rr' (C) t' t! > Haku 3 Figure 5 - Woman of Amendment f$ Procedural Amendment (Voluntary) 62.6. -1 Showa year month day each

Claims (3)

[Claims] (1) On the transmitting side, a discrete audio signal is input and divided into a predetermined first time interval, a pitch parameter representing pitch is extracted from the audio signal, and a second time interval shorter than the first time interval is input. and selecting one of the second time intervals to represent a spectral parameter representing a short-time spectral envelope of the audio signal in the selected interval and a sound source of the audio signal in the selected interval. The spectrum parameter and sound source information are combined and outputted with the pitch parameter, and the receiving side restores the second time interval based on the pitch parameter and information representing the sound source of the selected interval. and a spectral parameter of the selected section to synthesize the audio signal. (2) Divide the input audio signal into a predetermined first time interval, extract and encode a pitch parameter representing the pitch from the audio signal, and divide the input audio signal into a predetermined first time interval based on the pitch parameter. a dividing circuit for dividing into short second time intervals; and one of said second time intervals;
a parameter calculation circuit that selects a section and calculates and encodes a spectral parameter representing a short-time spectral envelope of the audio signal in the selected section and a pulse train representing a sound source of the audio signal in the selected section; A speech encoding device comprising: a multiplexer circuit that combines and outputs a code representing a sound source of the selected interval, a code representing a spectrum parameter of the selected interval, and a code representing the pitch parameter. (3) Input a code sequence in which a code representing the pitch parameter, a code representing the spectral parameter of the selected interval, and a code representing the sound source of the selected interval are combined for each predetermined first time interval. a demultiplexer circuit that separates and decodes a code representing the pitch parameter, a code representing the spectral parameter of the selected section, and a code representing sound source information of the selected section;
a dividing circuit that restores a second time interval shorter than the first time interval based on the decoded pitch parameter; and a division circuit that restores a second time interval that is shorter than the first time interval based on the decoded pitch parameter; a driving sound source signal restoring circuit that restores the driving sound source signal by performing processing to give a change in the driving sound source signal; and a driving sound source signal restoration circuit that synthesizes and outputs an audio signal using the decoded spectrum parameter of the selected section and the restored driving sound source signal. A speech decoding device comprising a synthesis filter circuit.