JP2560682B2 - Speech signal coding / decoding method and apparatus - Google Patents

Description

The present invention relates to a speech signal coding / decoding method and apparatus, and more particularly to a coding / decoding method for coding / decoding a speech signal with high quality at a low bit rate. A method and its apparatus.

(Prior Art and its Problems) As a method of encoding a voice signal at a transmission rate of about 16 Kbit / sec or less, a multi-pulse drive voice encoding method has been proposed. For more information on this method, see “Anew Model of LPC Excitement for Producing Natural Sounding” by BSATAL et al.
Speech at Low Bit Rate "(" A NEW
MODEL OF LPC EXCITATION FOR PRODUCING NATURAL−SOU
NDING SPEECH AT LOW BIT RATES ") (PRO.
ICASSP, pp 614-617, 1982) (reference 1) and so on, so that a brief description will be given here.

FIG. 3 is a block diagram showing the processing on the encoder side of the conventional method described in Document 1 above. In the figure, a discrete audio signal is input from the encoder input terminal 400 and accumulated for one frame in the buffer memory circuit 410. The subtractor 420 and the K parameter calculation circuit 480 receive the signal from the buffer memory circuit 410. However, in Reference 1 above, instead of the K parameter, the REFLECTION COE
FFICIENTS), which is the same as the K parameter. The K parameter calculation circuit 480 obtains a 16th order (1 <i <16) K parameter Ki representing the speech spectrum envelope in the frame according to the covariance method, and the synthesis filter 43
Output to 0. FIG. 4 shows an example of the pulse train d (n) in which the sound source pulse generation circuit 440 generates a predetermined number of pulse trains d (n) in one frame. In FIG. 4, the horizontal axis represents discrete time and the vertical axis represents amplitude.

Returning to FIG. 3, the synthesis filter 430 uses the pulse train d
Driven by (n), the synthesized voice signal (n) is obtained. (N) is output to the subtractor 420 and the original speech X
The difference signal e (n) from (n) is calculated. Weighting circuit
490 is a weighting error ew using the weighting function w (n)
Calculate (n).

Here, the characteristic of the weighting function w (n) is that the Z conversion value is W
If it is (Z), it is expressed by the following equation using the prediction coefficient value of the synthesis filter.

Here, r is a constant of 0 <r <1, and determines the frequency characteristic of W (Z). The reason why W (Z) is determined depending on the frequency characteristic of the synthesizing filter in the equation (1) is to use the audible masking effect that an error in the vicinity of the formant is hard to hear.

FIG. 5 shows a spectrum of an audio signal in a frame shown in FIG. 5 and an example of frequency characteristics of W (Z). Where r
= 0.8. In the figure, the horizontal axis is frequency (maximum 4KH
z), and the vertical axis shows the logarithmic amplitude (maximum 60 dB).
The upper curve shows the spectrum of the audio signal and the lower curve shows the W (Z) spectrum.

Returning to FIG. 3, the weighted error ew (n) is fed back to the error minimization circuit 450. The error minimization circuit 450 calculates the error power ε according to the following equation.

Here, N represents the number of samples, which is 40 samples (5 msec) in Document 1 above. Next, the error minimization circuit 450 uses the sound source pulse calculation circuit 440 to reduce the error power ε.
For pulse position and amplitude information. Synthesis filter
430 calculates the total signal (n) using this sound source pulse train as a driving source, and the subtracter 420 calculates the error e calculated previously.
The original synthesized signal (n) is subtracted from (n) and output to the weighting circuit 490. The weighting circuit 490 calculates a weighting error and feeds it back to the error minimizing circuit 450, and the error minimizing circuit 450 adjusts the sound source pulse so as to reduce the error power.

In this way, a series of processes from the generation of the sound source pulse train to the generation of the sound source pulse train by minimizing the error is repeated until the sound source pulse train reaches a predetermined value.

The information to be transmitted in the case of this method is K of the synthesis filter.
The parameter Ki (1 ≦ i ≦ 16) and the position and amplitude of the sound source pulse train are considered to be an effective method in the region where the transmission bit rate is 16 Kbps or less.

However, this conventional method has a drawback that the amount of calculation is very large. This is when calculating the source pulse
This is because the signals are once synthesized, the error power with respect to the original speech signal is calculated, and the pulse amplitude and position are adjusted so as to reduce it. In other words, the biggest problem is that the pulse calculation requires synthesis processing. Further, this conventional method has a drawback that the quality of synthesized speech is deteriorated for speech with a high pitch frequency when the transmission rate is lowered. This is because when the pitch frequency is high, there are many pitch waveforms in the frame for pulse calculation, and a large number of pulses are required to successfully combine them.

(Object of the Invention) It is an object of the present invention to provide a highly efficient speech signal coding / decoding method and apparatus capable of synthesizing a high quality speech even with a low transmission bit rate with a relatively small amount of calculation. It is in.

(Structure of the Invention) According to the speech signal coding / decoding method of the present invention, a discrete speech signal is input on the transmitting side, a pitch parameter representing a pitch is obtained from the speech signal, and each time section corresponding to the pitch parameter is obtained. A spectrum parameter representing a preset short-time spectrum envelope of the voice signal is divided into two, and a sound source pulse train for representing the voice signal is generated on the basis of the pitch parameter and the spectrum parameter. Is obtained and encoded for a part of a plurality of sections, and outputs a code in which a code representing the pitch parameter, a code representing the spectrum parameter, and a code representing the excitation pulse train are combined, Now enter the combined code,
A code representing the pitch parameter, a code representing the spectrum parameter, and a code representing the excitation pulse train are separated and decoded from the combined code, and the time interval is restored based on the decoded pitch parameter. Restore the driving excitation signal based on the decoded spectrum parameter and the decoded excitation pulse train,
The voice signal is synthesized with respect to the section obtained according to the time section.

A speech signal coding device of the present invention extracts a pitch calculation circuit for extracting and coding a pitch parameter representing a pitch from an input speech signal, and a spectrum parameter representing a preset short-time spectrum envelope of the speech signal. A parameter calculation circuit for encoding, a frame section setting circuit for obtaining a division position for dividing the audio signal based on the pitch parameter, and setting a frame section including a plurality of divided time sections; An autocorrelation function calculation circuit for calculating an autocorrelation sequence of an impulse response sequence according to a spectrum of time, a crosscorrelation function calculation circuit for calculating a crosscorrelation sequence of an impulse response sequence according to the short-time spectrum, and the self Source pulse for a part of the frame interval using the correlation sequence and the cross-correlation sequence A drive signal calculating and coding circuit for coding is calculated and output code and the drive signal calculating and the output code and the parameter calculation circuit of said pitch calculation circuit
And a multiplexer circuit that outputs a code sequence in which the output code of the encoding circuit is combined.

The speech signal decoding device of the present invention is input with a code sequence in which a code representing a pitch parameter, a code representing a spectrum parameter, and a code representing an excitation pulse train are input, and a code representing the pitch parameter and a code representing the spectrum parameter. And a code representing the excitation pulse train are separated and decoded, and a demultiplexer / restoring circuit that restores a time section corresponding to a pitch cycle based on the decoded pitch parameter, the restored time section, and the A driving sound source signal restoration circuit that restores a driving sound source signal based on a decoded sound source pulse train, and an audio signal of a section obtained according to the time section based on the driving sound source signal and the decoded spectrum parameter. And a synthesizing filter circuit for synthesizing.

(Principle of the Invention) The present invention utilizes the periodicity of a voice signal, and a transmitting side obtains a pitch parameter representing a pitch period from the voice signal,
The audio signal is divided into time intervals based on the pitch cycle. Then, a spectrum parameter representing a short-time spectrum envelope of the voice signal is obtained. A sound source pulse train for representing the voice signal based on the spectrum parameter is one of the time intervals included in the time interval obtained according to the time interval. It is calculated for each section and transmitted to the receiving side. Here, for calculation of the sound source pulse train, for example, “Japanese Patent Application No. 57-231606”
The formula (21) described in can be referred to. On the receiving side, a sound source signal is generated using the amplitude and position of the received sound source pulse train, and the drive sound source signal is restored by interpolating this sound source signal during the time interval when the sound source pulse train was not transmitted, and this drive sound source signal The speech signal is synthesized for the section obtained according to the time section using the spectrum parameter.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 (a) is a block diagram showing an embodiment of the transmitting side of the audio signal encoding device according to the present invention.
FIG. 6B is a block diagram showing an embodiment of the receiving side.

In FIG. 1A, the audio signal X (n) is input and stored in the buffer memory circuit 110 by a predetermined number of samples. The pitch analysis circuit 130 uses the output of the buffer memory circuit 110 to calculate the pitch period Pd. Pd
The calculation method of RVCO (RVCO
X) et al., "REAL-TIME IMPLEMENTAT"("REAL-TIMEIMPLEMENTAT")
ION OF TIME DOMAIN HARMONIC SCALING OF SPEECH SIGN
ALS ") (IEEE TRANS.ASSP, pp258-2
72, 1983) (reference 2) and the like can be used.

Pitch encoding circuit 150 quantizes and encodes pitch period Pd by a known method with a predetermined number of quantization bits, and outputs code ld to multiplexer 260. In addition, Pd ′ obtained by decoding is output to the frame section setting circuit 155 and the drive signal calculation circuit 220.

The frame section setting circuit 155 obtains a division position for dividing an audio signal for each pitch cycle Pd 'using the pitch cycle Pd', and sets the frame section so as to include some time sections which are further divided. Hereinafter, the divided time interval will be referred to as a subframe. Frame section is buffer memory circuit
It is output to 110, the synthesis filter circuit 250, and the drive signal restoration circuit 240. The division position is output to the drive signal calculation circuit 220.

Next, the K parameter calculation circuit 140 is the buffer memory circuit 1
A voice signal is input based on the frame interval from 10 and a K parameter Ki representing the spectral envelope of the input voice signal
Is calculated. Here, the K parameter is the same parameter as the PARCOR coefficient. The autocorrelation method is well known as a method for calculating the K parameter. For more information on this method, see “Quantization Properties of Transmission Parameters in Linear Predictive Systems (“ QUANTIZATION PROPERTIES OF ”by JOHN MAKHOUL and colleagues.
TRANSMISSION PARAMETERS IN LINEARPREDICTIVE SYSTE
MS ") paper (IEEE TRANS.ASSP, pp309-32
1,19753) (reference 3), etc., and therefore the description is omitted here.

Returning to FIG. 1A, the K parameter Ki is output to the K parameter encoding circuit 160. The K parameter encoding circuit 160 encodes Ki by a known method based on a predetermined number of quantization bits, and multiplexes the code li.
Output to 260. Further, the K parameter encoding circuit 160 is li
Is converted into a prediction coefficient value ai 'by a well-known method using the K parameter decoded value Ki' obtained by decoding the impulse response calculation circuit 170, the weighting circuit 200, and the synthesis filter circuit 250.
Output to.

The impulse response calculation circuit 170 inputs the prediction coefficient value ai ′ from the K parameter encoding circuit 160, and the impulse response hw (n) representing the transfer function of the weighted synthesis filter.
Is calculated. Here, the same method as the impulse response calculation circuit 210 described in FIG. 4 (a) of "Japanese Patent Application No. 59-042305" can be used for the calculation of hw (n). The impulse response hw (n) is output to the autocorrelation function calculation circuit 180 and the cross-correlation function calculation circuit 210.

The autocorrelation function calculation circuit 180 is an impulse response calculation circuit 1
The impulse response hw (n) is input from 70, and the autocorrelation function is calculated according to the following equation.

The autocorrelation function Rhh (m) is output to the drive signal calculation circuit 220.

Next, the subtractor 120 inputs the audio signal X (n) from the buffer memory circuit 110 by the number of samples based on the frame section in the frame section setting circuit 155, and outputs the output (n) of the synthesis filter circuit 250 from X (n). Is subtracted and the result e
(N) is output to the weighting circuit 200.

The weighting circuit 200 inputs e (n), and also inputs the prediction coefficient ai ′ from the K parameter encoding circuit 160,
Ew (n) obtained by weighting (n) is output. Here, for calculating ew (n), for example, “Japanese Patent Application No. 59-04
The same method as the weighting circuit 410 shown in FIG. 4 (a) of 2305 ″ can be used.

The cross-correlation function calculation circuit 210 outputs the weighting circuit 200 to ew
(N) is input, the impulse response hw (n) is input from the impulse response calculation circuit 170, and the cross-correlation function is calculated according to the following equation.
Calculate _hx (m).

The cross-correlation function _hx (m) is output to the drive signal calculation circuit 220.

Next, the drive signal calculation circuit 220 uses several source pulse trains for representing a voice signal using the cross-correlation function sequence and the auto-correlation function sequence divided using the division positions input from the frame period setting circuit 155. The calculation is performed for a part of the time section (subframe). This procedure is shown below. Here, as an example, for two subframes, a case where a sound source pulse train is obtained for one of the subframes and the sound source pulses are scattered for the remaining subframes will be described.

As an example, FIG. 2 (a) shows an audio waveform for four subframes, FIG. 2 (b) shows a cross-correlation function sequence obtained from the audio waveform of FIG. 2 (a), and FIG. 2 (c) shows 4 Each of the subframe sections is shown.

Next, one of the two sub-frames (here, the front sub-frame, and in FIG. 2 (c),
For a sub-frame indicated by 1 and 3), a predetermined number of sound source pulse trains are calculated. Here, for calculation of the sound source pulse train, for example, “Japanese Patent Application No. 57-231606”
The method represented by the formula (21) described in (4) can be used. FIG. 2D shows an example in which a predetermined number of excitation pulse trains are obtained for the first and third subframes. In FIG. 2 (d), the number of pulses assigned to each subframe is four.

The pulse amplitude and position thus obtained are output to the encoding circuit 230.

The encoding circuit 230 encodes the amplitude and position of the input pulse. Then, the amplitude of the pulse and the sign of the position are output to the multiplexer 260. Also, the amplitude of the input pulse,
The decoded values gi 'and mi' of the position are output to the drive signal restoration circuit 240. Here, as a pulse encoding method, for example, “Japanese Patent Application No.
The same method as the encoding circuit 250 described in −231605 ″ can be used. As other encoding methods, various methods such as a variable length encoding method or a differential encoding method can be considered.

The drive signal restoration circuit 240 uses the decoded values of the amplitude and position of the pulse input from the encoding circuit 230, and FIG.
Source pulse trains are generated for 1 and 3 subframes. Next, for subframes 2 and 4 in FIG. 2 (d),
Interpolation is performed based on the sound source pulse trains of subframes 1 and 3, respectively. As a special case of interpolation, for example, the sound source pulse trains of subframes 1 and 3 can be repeated with a pitch period Pd ′. Then, the frame section setting circuit 15
The drive sound source signal is obtained only for the frame section input from 5, and is output to the synthesis filter circuit 250.

The synthesis filter circuit 250 inputs the driving sound source signal and the K parameter and calculates the response signal (n) in one frame section. Here, in this calculation, for example, "Japanese Patent Application No. 57-23160"
The same method as the synthesis filter circuit 320 described in 5 "can be used.

The multiplexer circuit 260 is a K parameter encoding circuit 160.
Code lKi and pitch coding circuit 150 code ld and coding circuit
The code of 230 is input, accumulated once, and these are combined and output from the transmission side output terminal 270 at every predetermined time. This is the end of the description of the transmitting side of the audio signal encoding device according to the present invention.

Next, the configuration of the receiving side of the audio signal encoding device according to the present invention will be described with reference to FIG. 1 (b).

Of the codes input from the reception side input terminal 280 to the demultiplexer 290 at predetermined times, the code representing the K parameter, the code representing the pitch period, and the code representing the amplitude and position of the pulse are separated. , K parameter decoding circuit 330, pitch decoding circuit 320, and pulse decoding circuit 300, respectively.

The K parameter decoding circuit 330 decodes the K parameter and outputs the decoded value Ki ′ to the synthesis filter circuit 350.

The pitch decoding circuit 320 decodes the pitch period Pd ′ and outputs it to the drive signal restoration circuit 340 and the frame section setting circuit 345.

The pulse decoding circuit 300 decodes the pulse amplitude gi 'and the position mi' and outputs them to the drive signal restoring circuit 340.

The frame section setting circuit 345 performs the same operation as the frame section setting circuit 155 on the transmitting side, and outputs the obtained frame section to the drive signal calculation circuit 340 and the synthesis filter circuit 350.

The drive signal restoration circuit 340 is the drive signal restoration circuit 240 on the transmission side.
The same operation as the above is performed to obtain a driving sound source signal in one frame section and output it to the synthesis filter circuit 350.

The synthesis filter circuit 350 receives the driving sound source signal, performs the same operation as the synthesis filter circuit 250 on the transmitting side, calculates a synthesized voice signal (n) in one frame section, and outputs it to the buffer memory circuit 355.

The buffer memory circuit 355 stores the synthesized voice signal for a predetermined number of samples and then outputs it to the receiving side output terminal.
Output from 360.

This is the end of the description of the receiving side of the audio signal encoding device according to the present invention.

As a pulse calculation method in the drive signal calculation circuit 220, various methods can be used in addition to the method described in this embodiment. For example, a method of adjusting the amplitude of a pulse obtained in the past every time one pulse is obtained can be used. For details of this method, see a paper entitled “Study on Source Pulse Search Method in Multi-pulse Driven Speech Coding” by Ono et al. (Proceedings of the Acoustical Society of Japan 157,198).
3) Since it is described in (Reference 4) and the like, the description is omitted here.

Further, when obtaining the sound source pulse train in the drive signal calculation circuit 220, the pulse was obtained for each subframe, but a predetermined number of pulses are obtained for the entire section including some subframes. Good. Furthermore, on the transmitting side of the present embodiment, the value of the K parameter is constant in the subframes belonging to the frame section used for the analysis of the K parameter, but the pulse is generated while smoothly changing the value of the K parameter for each subframe. You may ask. Specifically, the value of the K parameter is interpolated for each subframe using the values of the K parameter of the preceding and succeeding frames, this value is converted into a prediction coefficient, and the weighting circuit 200, the impulse response calculation circuit 170, The pulse is calculated using the cross-correlation function and autocorrelation function that are output to the synthesis filter circuit 250 and updated by updating the coefficient for each subframe. In this way, better voice can be synthesized. In this case, a method other than linear interpolation can be considered as an interpolation method for the parameters of the synthesis filter.

When interpolating the parameters of the synthesis filter, K
In addition to the method of interpolating the parameters, for example, the method of interpolating the prediction coefficient (however, in this case it is necessary to check the stability of the filter), the method of interpolating the logarithmic cross-section function, and the autocorrelation function An interpolation method or the like can also be used. These specific methods are
"SPACH ANALYSIS AND SYNTHESIS BY LINEAR PREDICTION OF THE SPEECH WAVE" by BSATAL et al.
SIS AND SYNTHESIS BY LINEAR PREDICTION OF THE SPEE
CH WAVE ") (J.ACOUST.SOC.AM., Pp637-
655, 1971) (reference 5) and the like, and therefore description thereof will be omitted.

Further, in the present embodiment, the subtraction, the cross-correlation function calculation, and the sound source pulse calculation processing are performed on the voice signal of the frame section including some sub-frame sections, but the above processing is performed for each sub-frame section. May be. Furthermore, K
The frame section used for the parameter analysis may be the same as or different from the section in which the subtraction cross-correlation function calculation and the sound source pulse calculation process are performed. As for the frame section used for the analysis of the K parameter, the frame section may be set to be short in the voice changing section and the frame section may be set to be long in the stationary section. By doing so, it is possible to improve the quality at the voice change portion and reduce the transmission bit rate. In such a configuration, the K parameter analysis frame section may be determined according to the pitch cycle or may be determined regardless of the pitch cycle.

Also, an example in which a pulse for one subframe is transmitted for two subframes has been described, but a pulse for one subframe may be transmitted for three or more subframes. This makes it possible to further reduce the transmission bit rate. The thinning rate does not have to be constant. For example, it can be changed according to the pitch period. That is, the number of subframes to be thinned out may be increased when the pitch period is short, and the number of thinned out subframes may be reduced when the pitch period is long.

As another configuration of the present invention, the following can be performed. First, the transmitter side first obtains the excitation position of the sound source pulse. For this purpose, the drive signal calculation circuit 220 may find the first sound source pulse. Next, the start position of the time section corresponding to the pitch cycle is obtained based on this excitation position, and the audio signal is divided from this start position by the predetermined number of sections for each pitch cycle. Then, a sound source pulse train that satisfactorily represents a voice signal may be obtained for some of some time intervals using the same method as in the above-described embodiment. However, in the case of such a configuration, it is necessary to transmit the information on the start position of the time section to the receiving side.

Note that, as is well known in the field of digital signal processing, the autocorrelation function can be calculated from a power spectrum. Also, the cross-correlation function can be calculated from the cross-power spectrum. Regarding these correspondences, see "Digital signal processing" and "DIGITAL SIGNAL PROC" by AVOPPENHEIM and others.
Since it has been described in detail in Chapter 8 such as a book titled "ESSING" (reference 6), its description is omitted here.

(Effects of the present invention) As described above, the present invention utilizes the periodicity of the audio signal on the transmission side, divides the audio signal into time intervals according to the pitch cycle, and sets a predetermined short time of the audio signal. Based on the spectrum parameter that represents the spectrum envelope, the sound source pulse train that represents the voice signal is obtained for some of the multiple time intervals, encoded, and output, and the receiving side separates the code that represents the received pitch parameter. , The time interval corresponding to the pitch period restored based on the decoded pitch parameter and the driving source signal restored based on the decoded source pulse train, and the separated and decoded spectral parameters Since the audio signal of the section obtained according to the time section is synthesized on the basis of the This has the effect of being able to synthesize quality voice.

[Brief description of drawings]

1 (a) and 1 (b) are block diagrams showing an embodiment of a high-efficiency voice signal conversion and decoding apparatus according to the present invention, and FIG. 2 is a diagram showing an example of processing contents in the drive signal calculation circuit 220. 3, FIG. 3 is a block diagram showing a configuration of a conventional system, FIG. 4 is a diagram showing an example of a sound source signal represented by multi-pulses, and FIG. 5 is a frequency characteristic of an audio signal and weighting shown in FIG. It is a figure which shows an example of the characteristic of a circuit. In the figure, 110,355,410 ... buffer memory circuit, 12
0,420 …… subtraction circuit, 250,350,430 …… synthesis filter circuit, 200,490 …… weighting circuit, 170 …… impulse response calculation circuit, 180 …… autocorrelation function calculation circuit, 210 …… cross-correlation function calculation circuit, 220 …… Drive signal calculation circuit, 240,340
...... Drive signal restoration circuit, 130 …… Pitch analysis circuit, 140,
480 ... K parameter calculation circuit, 150 ... pitch encoding circuit, 155,345 ... frame section setting circuit, 160 ... K parameter encoding circuit, 230 ... encoding circuit, 260 ... multiplexer, 290 ... demultiplexer , 300 ... pulse decoding circuit, 320 ... pitch decoding circuit, 330 ... K parameter decoding circuit, 440 ... excitation pulse generation circuit, 450 ... error minimization circuit, respectively.

Claims (3)

(57) [Claims] 1. A transmission side inputs a discrete voice signal, obtains a pitch parameter representing a pitch from the voice signal, divides the voice signal for each time interval according to the pitch parameter, and pre-generates the voice signal. A spectrum parameter representing a set short-time spectrum envelope is obtained, and a sound source pulse train for representing the voice signal is generated for a part of a plurality of the time intervals based on the pitch parameter and the spectrum parameter. Then, a code representing the pitch parameter, a code representing the spectrum parameter, and a code representing the excitation pulse train is output in combination, and the reception side inputs the combined code, A code representing the pitch parameter and a spectrum parameter from the combined code Signal and a code representing the excitation pulse train are separated and decoded, and the time interval is restored based on the decoded pitch parameter to obtain the decoded spectrum parameter and the decoded excitation pulse train. A voice signal encoding / decoding method, characterized in that the drive excitation signal is restored to, and the voice signal is synthesized with respect to a section obtained according to the time section. 2. A pitch calculation circuit for extracting and encoding a pitch parameter representing a pitch from an input voice signal, and a parameter calculation for extracting and encoding a spectrum parameter representing a preset short-time spectrum envelope of the voice signal. A circuit, a frame section setting circuit for determining a division position for dividing the audio signal based on the pitch parameter, and setting a frame section including a plurality of divided time sections, and a frame section setting circuit An autocorrelation function calculating circuit for calculating an autocorrelation sequence of the impulse response sequence, a crosscorrelation function calculating circuit for calculating a crosscorrelation sequence of the impulse response sequence according to the short-time spectrum, the autocorrelation sequence and the mutual correlation The excitation pulse train is calculated and encoded for a part of the frame interval using the correlation sequence A drive signal calculation / encoding circuit, and a multiplexer circuit that outputs a code sequence that combines the output code of the pitch calculation circuit, the output code of the parameter calculation circuit, and the output code of the drive signal calculation / encoding circuit. An audio signal encoding device having. 3. A code sequence in which a code representing a pitch parameter, a code representing a spectrum parameter, and a code representing an excitation pulse train are combined and inputted, a code representing the pitch parameter, a code representing the spectrum parameter, and the excitation pulse train. A demultiplexer / restoring circuit that restores the time interval corresponding to the pitch period based on the decoded pitch parameter by decoding the code representing A driving sound source signal restoration circuit for restoring a driving sound source signal based on a pulse train, and a synthesis filter for synthesizing an audio signal in a section obtained according to the time section based on the driving sound source signal and the decoded spectrum parameter An audio signal decoding device comprising: a circuit.