JP2615548B2 - Highly efficient speech coding system and its device. - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an encoding method and apparatus for encoding an audio signal with high quality at a low bit rate.

(Prior Art) A vocoder (VOCODER) is known as a method for encoding an audio signal at a low transmission bit rate (for example, about 4.8 kbps). The principle of this method is described in, for example, "VOCODERS: ANALYSIS AND SYNTHEN" by MR SCROEDER.
SIS OF SPEECH ") (PROC.IEEE, pp720-7
34, MAY, 1966) (Literature 1).
LPCVOCODER (LPCVOCODER) is known as a vocoder using the linear prediction analysis method, and its contents are described, for example, by JD Markel (JDMA).
RKEL) et al. “A Linear Prediction
Vocoder based upon the autocorrelation method ”(“ ALINEAR PREDICTION VOCODER BASED U
PON THE AUTOCORRELATION METHOD ") (IEE
E TRANS.ASSP, pp124-134, APRIL, 1974) (Reference 2) and the like. The present invention is an improvement of the sound source section of the VOCODER, and has a close relationship with the LPCVOCODER. Therefore, the LPCVOCODER will be briefly described below with a focus on the configuration of the synthesis section.

FIG. 4 is a block diagram showing a synthesis unit (reception unit) of LPCVOCODER described in Reference 2. The synthesis unit is the sound source generation unit 500
And a synthesis filter 510. The sound source generator 500 includes an impulse generator 501, a noise generator 502, and a voiced / unvoiced switching circuit.
503 and a gain circuit 504. In VOCODER,
The audio signal is voiced and unvoiced every short time (for example, 20 msec).
In the case of voice, the impulse generator 501 generates a pulse train having a time interval of the pitch period Pd. On the other hand, when there is no voice, white noise is generated from the noise generator 502. The voiced / unvoiced control is performed by the switching circuit 503.
It is done in. The signal generated in this way is given a gain G by a gain circuit 504 and output to the synthesis filter 510 as a sound source signal d (n).

The synthesis filter 510 synthesizes and outputs the sound x (n) using the sound source signal d (n) and the filter parameter Ki. Here, pitch period Pd, voiced / unvoiced switching signal (V / UV),
The gain G and the filter parameter Ki are calculated at predetermined intervals on the analysis side (transmission side) and transmitted to the reception side.

(Problems to be Solved by the Invention) In the LPCVOCODER described above, the transmission information is a pitch period, voiced / unvoiced voice, gain, and filter parameters, and a voice signal can be synthesized from these information. Can be reduced (for example, about 4.5 kbps). However, it has been difficult to synthesize a high-quality voice by the conventional method. That is, when the sound source signal is voiced, the sound source is represented by one impulse per pitch and does not include phase information.
Naturalness was considerably impaired, and the synthesized sound was a so-called mechanical sound. Also, voice is called voiced and unvoiced.
Since the impulse sound source is switched to the noise source, the sound source is switched to the noise source, so that there is a defect that when a voiced / unvoiced discrimination error occurs, a large quality deterioration is caused. In addition, the sound source could not be well represented at the switching part between unvoiced and voiced, and deterioration occurred. Further, when the pitch period is determined to be shifted, there is a disadvantage that a great quality deterioration is caused.

As a method for improving the sound source, for example, Japanese Patent Application No. 59-272435
As described in the specification (Document 3) and the like, a method of expressing a sound source by a combination of pulse trains and transmitting a pulse train in a representative pitch section is known. In this method, the above problem can be improved and a good quality can be obtained in a voiced section in which the pitch period is clear, but a voiceless section in which the pitch is not clear and the sound source is noisy, and an unvoiced section and a voiced section are used. In the transient part, when the transmission bit rate is low, the sound source cannot be displayed well, and the quality deteriorates.

SUMMARY OF THE INVENTION An object of the present invention is to provide a high-efficiency speech coding system and a device capable of synthesizing high-quality speech with a relatively small amount of computation even at a low transmission bit rate of about 4.8 kbps.

(Means for Solving the Problem) According to the high-efficiency speech coding method of the present invention, a discrete speech signal is inputted on the transmitting side, divided into predetermined time intervals, and a short-time spectrum envelope is obtained from the speech signal. Extracting a spectral parameter and a pitch parameter representing a pitch cycle, dividing the audio signal at the time interval into small sections corresponding to the pitch cycle, and representing a sound source for representing the audio signal as a representative of the small sections. Expressed by a combination of a pulse train or a pulse train and noise in a typical section, combining the information representing the sound source, the pitch parameter and the spectrum parameter, and outputting the combined data.On the receiving side, based on the information representing the pitch parameter and the sound source. Restoring the drive sound source signal by performing processing to give a temporally smooth change to the pulse train of the representative section for each pitch cycle And synthesizing the audio signal using the spectrum parameter.

Further, the encoding device of the present invention divides an input audio signal into predetermined time intervals, extracts and encodes a spectrum parameter representing a short-time spectrum envelope and a pitch parameter representing a pitch period from the audio signal. A parameter calculation circuit, the audio signal at the time interval is divided into small sections corresponding to the pitch period, and a sound source for representing the audio signal is a pulse train or a pulse train of a representative section of the small section and noise. A drive signal calculation circuit for obtaining a combination and encoding information representing the sound source; and a multiplexer circuit for combining and outputting an output code of the parameter calculation circuit and an output code of the drive signal calculation circuit. .

Further, the decoding device of the present invention, a code representing a combination of a code representing a pitch parameter, a code representing a spectrum parameter, and a code representing excitation information is input, and a code representing the pitch parameter and a code representing the spectrum parameter. A demultiplexer circuit that separates and decodes the code representing the excitation information, and a temporally smooth pulse train of a representative section based on the decoded pitch parameter and the decoded excitation information for each pitch cycle. When a pulse train and noise are used as a sound source, a driving sound source signal restoring circuit that generates a pulse train and noise based on the sound source information to restore a driving sound source signal, A synthesis filter circuit for synthesizing and outputting the audio signal based on the decoded spectral parameters.

(Operation) The present invention utilizes a periodicity of a sound signal to generate a sound source signal represented by a pulse train of a typical one-pitch section as described in Reference 3 and a sound source signal by a combination of a pulse and a noise source. Among them, a sound source signal capable of better expressing a voice signal is selected. As a method for obtaining a typical pulse train in one pitch section, see the aforementioned reference 3.
Can be used. As a method for obtaining the amplitude and the position of the pulse train, in addition to the method described in the aforementioned reference 3, for example, analysis-by-synthesis (AN
A method using the ALYSIS-by-SYNTHESIS (Abs) method is known, and details thereof are described in “A New Model of LPC Excitation for Producing Natural” by BSATAL et al. Sounding Speech At Low Bit Rate's ("A NEW MODEL
OF LPC EXCITATION FOR PRODUCING NATURAL SOUND
ING SPEECH ATLOW BIT RATES ") (PROC.I.
CASSP, pp. 614-617, 1982) (Reference 4).

On the other hand, in a method of obtaining a sound source by a combination of a pulse and noise, a predetermined number of pulse trains are obtained for the entire frame, and then the amplitude and phase of the noise source are calculated.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 (a) is a block diagram showing one embodiment of a transmitting side of the high-efficiency speech coding system according to the present invention, and FIG. 1 (b) is a block diagram showing one embodiment of a receiving side.

In FIG. 1A, an audio signal X (n) is input and stored in a buffer memory circuit 110 for a predetermined number of samples. Next, the K parameter calculation circuit 140
An audio signal having a predetermined number of samples is input from the buffer memory circuit 110, and a K parameter representing a spectrum envelope of the audio signal is calculated. Here, the K parameter is the same parameter as the PARCOR coefficient. As a calculation method of the K parameter, an autocorrelation method is well known. For more information on this method, see John Makoul (JOHN M
AKHOUL et al., “Quantization Properties of Transmission Parameters in Linear Predictive Systems (“ QUANTIZA
TION PROPERTIES OF TRANSMISSION PARAMETERS IN
LINEAR PREDICTIVE SYSTEMS ") (IEEE T
RANS.ASSP, pp. 309-321, 1983) (Reference 5) and the like, and a description thereof is omitted here. Returning to FIG. 1 (a), K parameters K _i is output to the K parameter coding circuit 160. K parameter encoding circuit 160
Is, K _i on the basis of the number of quantization bits which is predetermined
And outputs the code l _i to the multiplexer 260.
Further, the K parameter encoding circuit 160 uses the K parameter decoded value K _i ′ obtained by decoding l _i , converts it into a prediction coefficient value a _i ′, and converts the impulse response calculation circuit 170 and the weighting circuit 200
And output to Also, the K parameter decoded value K _i '
Output to 255.

Pitch analysis circuit 130 calculates the pitch period P _d using the output of the buffer memory circuit 110. The calculation method of P _d is described in, for example, “Real-time Implementation of Time Domain Harmonic Scaling of RVCOX” et al.
Speech signals ”(“ REAL-TIME IMPLEMENTATIO
N OF TIME DOMAIN HARMONIC SCALING OF SPEECH SIG
NALS ") (IEEE TRANS.ASSP, pp258-
272, 1983) (Reference 6) and the like.

Pitch encoding circuit 150 quantizes encoded with the number of quantization bits determined pitch period P _d advance, outputs the code l _d to the multiplexer 260. Also obtained by decoding
P _d ′ is output to the drive signal calculation circuit 220.

The impulse response calculation circuit 170 receives the prediction coefficient value a _i ′ from the K parameter encoding circuit 160, and receives an impulse response hw representing a transfer function of the weighted synthesis filter.
Calculate (n). Here, for the calculation of h _w (n), for example, FIG. 4 (a) of Japanese Patent Application No. 59-042305 (Reference 7) is used.
The same method as the impulse response calculation circuit 210 described in (1) can be used. The impulse response h _w (n) is output to the autocorrelation function calculation circuit 180 and the cross-correlation calculation circuit 210.

The auto-correlation function calculation circuit 180 receives the impulse response h _w (n) from the impulse response calculation circuit 170, calculates an auto-correlation function R _hh (m), and outputs it to the drive signal calculation circuit 220. Here, for the calculation of R _hh (m), for example, the same method as that of the autocorrelation function calculation circuit 180 described in Reference 7 can be used.

Next, the subtracter 120 subtracts the output of the synthesis filter circuit 250 for one frame from the audio signal X (n) of the buffer memory circuit 110, and outputs the result e (n) to the weighting circuit 200. The weighting circuit 200 receives e (n),
The prediction coefficient a _i ′ is input from the parameter coding circuit 160,
subjected to heavy moisture to the e (n) to output a e _w (n) was determined. Here, for the calculation of e (n), for example, the same method as that of the weighting circuit 410 described in FIG.

The cross-correlation function calculation circuit 210 calculates e _w
(N) is input, the impulse response h _w (n) is input from the impulse response calculation circuit 170, the cross-correlation function Ψ _hx (m) is calculated, and output to the drive signal calculation circuit 220. Where Ψ _hx
For the calculation of (m), for example, the same method as the cross-correlation function calculation circuit 210 described in Reference 7 can be used.

Next, the drive signal calculation circuit 220 first calculates a pulse train of a representative pitch section as a sound source signal representing a sound signal. Next, a sound source signal based on the pulse train and the noise source is calculated, and a sound source signal that can better represent the voice signal is selected from these. The determination of whether the pitch is clearly, it is possible to use a pitch gain P _g is a simplified method.

A method for obtaining the sound source signal will be described below. As a method of calculating a pulse train in a representative pitch section, for example, the same method as that of the drive signal calculation circuit 220 described in Reference 3 can be used. Therefore, only a brief description will be given here.

First, the frame is divided into subframes for each pitch period _Pd '. For this division, it is necessary to know the excitation position of the pitch, which can be known by obtaining a pulse train representing the sound source. That is, the excitation position of the pitch can be known from the position of the pulse obtained first. Here, for the calculation of the pulse cormorant, for example,
The method represented by the formula (21) described in the specification of 231606 (Reference 8) can be used. FIG. 2A shows the speech waveform of one frame, and FIG. 2B shows the first pulse obtained.
The state of a subframe divided using g1 and the position of this pulse is shown. Next, a predetermined number of pulses are calculated for each subframe. As a method of selecting a pitch section, for example, a method in which a subframe near the center of a frame is set as a representative pitch section and pulses included in this section are set as representative pulses can be considered. FIG. 2C shows the representative pitch pulse obtained in this manner. The amplitude and position of the pulse train in the representative pitch section are output to encoder 230. The subframe phase T and the subframe number of the representative pitch section (3 in FIG. 2C) are encoded with a predetermined number of bits as the representative pitch position, and output to the multiplexer 260.

Next, a method of obtaining a sound source using pulses and noise will be described. First, a predetermined number L of pulses is obtained for the entire frame using the above-described method. The signal (n) is synthesized using these pulses, and a signal X '(n) obtained by subtracting the synthesized signal (n) from the original audio signal X (n) is obtained.
The noise source is chosen so as to better represent (n). The specific method of this calculation will be described below. Now, let q (n) be the noise source,
Let G be the amplitude of the noise source and h be the impulse response of the synthesis filter.
Assuming that (n), the error power ε between the signal (n) synthesized from the noise source and the signal X ′ (n) can be expressed by the following equation.

The noise source amplitude G _x can be determined so as to minimize the above equation.

More specifically, only predetermined types (for example, B types) of noise source patterns are stored in the noise memory 225, and the noise sources are read out from the noise memory 225 one by one.
Then, the optimum amplitude G is obtained based on the equation (1b), and the error power at this time is calculated. Then, the above processing is repeated for the types of noise sources (B types) to find the type of noise source that minimizes the error power. The sound source calculation processing described above is particularly effective when the characteristics of the sound source change little by little in the transition section between the unvoiced section and the voiced section.

From the two types of sound sources obtained as described above, a sound source that reduces the error power between the audio signal and the synthesized signal is selected, and sound source information representing this sound source is output to the encoding circuit 230.

The encoding circuit 230, when a pulse train is input,
Encode the amplitude and position of the pulse train. Then, the amplitude and the sign of the position of the pulse train are output to the multiplexer 260.
In addition, the decoded values g _i ′ and m _i ′ of the pulse train amplitude and position are output to the drive signal restoration circuit 240. Here, as a pulse encoding method, for example, the same method as the encoding circuit 250 described in the above-mentioned reference 8 can be used.

When the information of the pulse and the noise source is input, the pulse train is encoded using the same method as described above,
For a noise source, a code representing the amplitude and the type of noise is encoded with a predetermined number of bits, and the code is output to a multiplexer. Further, it outputs the decoded value to drive signal restoration circuit 240.

The drive signal restoration circuit 240 generates an excitation signal for one frame using the decoded value input from the encoding circuit 230,
This is output to the synthesis filter circuit 250 as a driving sound source signal.

When a pulse train is used as the sound source, the interpolation circuit 255 inputs the pitch period P _d ′, the subframe phase T, and the representative pitch position, and interpolates the K parameter for the subframe divided for each pitch period P _d ′. I do. Here, the interpolation is linear interpolation, and the interpolation is performed using the values of the K parameter one frame before and one frame ahead. This interpolation is shown in FIG. In the figure, the i-th K in the j-th frame
As the parameter K _{i, j} , interpolation is performed for each subframe using the value K _{i, j−1} one frame before and the value K _{i, j + 1} one frame ahead. The K parameter obtained by interpolation in this way is output to the synthesis filter circuit 250.

When a pulse and noise are used as a sound source, interpolation is performed for each predetermined sample section. The interpolated K parameter is output to the synthesis filter circuit 250. The synthesis filter circuit 250 receives the driving sound source signal and the interpolated K parameter, and calculates a response signal (n) for one frame. Here, for this calculation, for example, the same method as that of the synthesis filter circuit 320 described in Reference 8 can be used.

The multiplexer circuit 260 is a K-parameter encoding circuit 1
The code l _{ki of} 60, the code l _d of the pitch coding circuit 150 and the code of the coding circuit 230 are input, and if a pulse train is used, the subframe phase and the representative pitch position are further input, and these are combined and transmitted. Output from the side output terminal 270. This concludes the description of the transmitting side of the high-efficiency speech coding system according to the present invention.

Next, the configuration of the receiving side of the speech coding system according to the present invention will be described with reference to FIG.

The demultiplexer 290 separates a code representing the K parameter, a code representing the pitch period, and a code representing the excitation information among the codes input from the receiving side input terminal 280, and separates them into the K parameter decoding circuit 330, Output to the decoding circuit 320 and the decoding circuit 300.

The K parameter decoding circuit 330 decodes the K parameter and outputs a decoded value K _i ′ to the interpolation circuit 335.

The pitch decoding circuit 320 decodes the pitch period P _d ′,
The signal is output to the drive signal restoration circuit 340 and the interpolation circuit 335.

The decoding circuit 300 decodes the sound source information and restores the drive signal
Output to

The drive signal restoring circuit 340 uses the pitch period decoded value P _d ′ to determine that a pulse train is used as a sound source if this is a value other than 0, and outputs a code representing a subframe phase and a representative pitch position as sound source information. , And are used to divide the frame into subframes for each pitch period P _d ′. Then, a pulse having an amplitude g 'is generated at a position m' for the subframe section represented by the representative pitch position. Next, a pulse is interpolated and determined for each sub-frame using the representative pitch pulse and the representative pulse one frame before and one frame ahead. In this manner, a pulse is generated for the entire one frame to restore the driving sound source signal and output to the synthesis filter circuit 350.

On the other hand, when the pulse and the noise are used as the sound source, the code representing the amplitude of the pulse train, the position and the amplitude and the type of the noise source are separated from the sound source information and decoded. For noise sources,
With respect to the noise memory 310 in which the same noise as that of the noise memory circuit 225 on the transmission side is stored, the decoded type is read as a read start position, a noise signal is read out for a predetermined number of samples, and the noise signal is multiplied by the amplitude G. Play the sound source. Now, assuming that the sample value of the noise signal is qi (n), the sound source signal v (n) can be expressed by the following equation.

V (n) = G · qi (n) (2) In the above equation, i indicates the type of the noise signal stored in the noise memory 310.

The driving pulse signal is restored by adding the decoded pulse train to the excitation signal of the above formula, and output to the synthesis filter circuit 350.

The interpolation circuit 335 performs the same operation as the interpolation circuit 225 on the transmission side, interpolates the decoded K parameter for each pitch period, and outputs the interpolated K parameter to the synthesis filter circuit 350.

The synthesis filter circuit 350 inputs the driving sound source signal and the interpolated K parameter, and outputs the synthesis filter circuit 250 on the transmission side.
Performs the same operation as that of the synthesized voice signal X for one frame.
(N) is calculated and output from the receiving-side output terminal 360.

The receiving side of the high-efficiency speech coding system according to the present invention has been described above.

In the drive signal calculation circuit 220, in order to satisfactorily represent various voices in the unvoiced section and realize a good transition between the unvoiced section and the voice section, when the sound source is represented by pulses and noise, the number of pulses May be adaptively changed from zero to several. In this case, it is necessary to transmit information indicating the number of pulses (for example, about 2 bits per frame). As a method of reducing the amount of calculation, for example, a pitch encoding circuit obtains a pitch gain from the value of an autocorrelation function separated by one pitch, and determines whether voiced or unvoiced on the transmitting side before calculating a sound source signal based on the magnitude of the pitch gain. Alternatively, a pulse train of a representative pitch section may be used as a sound source signal when voiced, and a combination of noise and a pulse train may be used when voiceless. In addition, as a method of determining voiced / unvoiced, other well-known methods can be used.

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 9) and the like, the description is omitted here.

In addition, when a pulse train was obtained by the drive signal calculation circuit 220, the frame was divided into sub-frames, and then a pulse train was obtained for each sub-frame. A given number of pulses may be obtained, and the pulses in the subframe may be used.

On the other hand, as another method of calculating the noise source, for example,
There is known a method of generating a random noise signal according to a Gaussian distribution for each subframe, and selecting noise that minimizes error power between a signal synthesized from the noise signal and a speech signal in a subframe section. For more information on this method, see “Stochastic Coding of Speech Signals at Berry Low Bit Rate” by BSATAL et al.
(“STOCHASTIC CODING OF SPEECH SIGNALS AT VERY LO
W BIT RATES ") (PROC., ICC84, pp. 1610-1)
613, 1984) (Literature 10). As another method, there is a method in which a predetermined number of samples are prepared as a single noise source, and the amplitude and phase (readout position) of the noise source are obtained from the prediction residual signal obtained by predicting the audio signal. Are known. In this method, the calculation is performed on the prediction residual, so that the amount of calculation can be reduced. For a detailed description of this method, see a paper entitled “Linear predictive analysis and synthesis using a driving sound source with residual modeled by noise” by Oyama (Proceedings of the Acoustical Society of Japan, Oct. 1984.
Pp. 165-166) (Reference 11). In an unvoiced section where the characteristics of the sound source are almost constant,
As described above, only the amplitude may be transmitted using a fixed noise source, and the amplitude and type of the noise source may be transmitted in the transient part. Furthermore,
The noise source may always be fixed in the unvoiced section.

On the transmitting side of the present embodiment, when obtaining a pulse for each subframe in a frame in a voiced section, the value of the K parameter is constant within the frame (that is, the characteristics of the synthesis filter do not change within the frame). However, the pulse may be obtained while smoothly changing the value of the K parameter for each subframe. Specifically, the value of the K parameter is interpolated for each subframe using the value of the K parameter of the preceding and succeeding frames, and this value is converted into a prediction coefficient, and the weighting circuit 200, the impulse response calculation circuit 170, A pulse is calculated using the cross-correlation function and the auto-correlation function which are output to the synthesis filter circuit 250 and updated for each sub-frame. In this way, a temporally smooth spectrum change can be obtained, and higher quality speech can be synthesized. In addition, when interpolating the values of the pulse and the K parameter, interpolation may be performed in synchronization with the pitch cycle based on a representative pitch section, or one or both of the pulse and the K parameter may be determined in advance. The interpolation may be performed based on a determined pitch section (for example, a pitch section near the center of the frame). When both use such an interpolation method, the code representing the position of the representative pitch section does not need to be transmitted, and the transmission bit rate can be reduced. On the other hand, a method of interpolating the pulse and the K parameter without synchronizing with the pitch period is also conceivable. In this case, the frame is divided into predetermined time intervals (for example, about 2.5 msec), and interpolation processing is performed for each section. In this case, the transmission bit rate can be reduced because the subframe phase does not need to be transmitted. In this case, as the interpolation reference section, a representative section may be searched for on the transmitting side or may be determined in advance (for example, near the center of the frame). In the latter case, the subframe phase and the representative pitch position need not be transmitted, and the bit rate can be further reduced.

As a method of reducing the calculation amount, the K parameter interpolation processing may be performed only on the receiving side. By doing so, the interpolation circuit 255 on the transmission side can be omitted.

Further, as a method of selecting the representative pitch section, another method such as a method of selecting a subframe including a pulse having a large absolute value can be used. In addition, a section in which good sound can be reproduced can be searched for each frame. Although the pitch period is fixed when subframe division is performed, this value may be interpolated using the pitch periods of the preceding and succeeding frames. In this case, the change in the pitch period becomes smoother in time, and a better sound can be obtained.

Next, as a method of interpolating the pulse, the parameters of the synthesis filter, and the pitch period, a method other than linear interpolation may be used. For example, logarithmic interpolation or the like can be considered for the pulse and pitch period. In the present embodiment, when the parameters of the synthesis filter are interpolated, the K parameter is interpolated. However, for example, a prediction coefficient (however, in this case, it is necessary to check the stability of the filter), a logarithmic cross section function,
A method of interpolating a formant parameter or an autocorrelation function may be used. These specific methods are described in a speech by BSATAL and others.
Analysis and Synthesis by Linear Prediction of the Speech Wave ”
(“SPEECH ANALYSIS AND SYNTHESIS BY LINEAR PREDIC
TION OF THE SPEECH WAVE ") (J. ACOUST.S
OC.AM., pp637-655, 1971) (Reference 12) and the like, and a description thereof will not be repeated.

In the present embodiment, the analysis of the K parameter and the calculation of the sound source pulse train are performed with the frame length being fixed, but the frame length may be variable. In such a case, the frame length can be shortened in the voice change section and the frame length can be increased in the stationary section, so that the transmission bit rate can be reduced. Furthermore, if the frame length is determined according to the pitch period (for example, an integral multiple of the pitch period), the sub-frame phase described in the present embodiment does not need to be transmitted, so that the transmission bit rate can be further reduced. . As another configuration method of the present invention, the drive signal restoration circuit 240, the synthesis filter circuit 250, the interpolation circuit 255, and the subtraction circuit 1 in FIG.
A configuration in which 20 is omitted can also be adopted. In such a case, it is not necessary to synthesize the audio signal on the transmission 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 be calculated from a power spectrum. Also, the cross-correlation function can be calculated from the cross-power spectrum. For information on these correspondences, see “Digital Signal Processing” and “DIGITAL SIGNAL PRO” by AVOPPENHEIM and others.
This is described in detail in Chapter 8, such as a book entitled "CESSING" (Reference 13), and will not be described here.

(Effects of the Present Invention) As described above, according to the present invention, as a sound source signal,
Since a sound source that can reproduce an audio signal better is selected from a sound source based on a pulse train of a representative one-pitch interval using the periodicity of the audio signal and a sound source based on a combination of pulses and noise, a low transmission bit is selected. Also in the late, there is an effect that high-quality speech can be synthesized regardless of a voiced section, an unvoiced section, and a transition section between the unvoiced section and the voiced section.

[Brief description of the drawings]

1 (a) and 1 (b) are block diagrams showing an embodiment of a high-efficiency speech coding system according to the present invention, FIG. 2 is a diagram showing an example of processing contents in a drive signal calculation circuit 220, and FIG.
FIG. 4 is a diagram showing a processing example of the interpolation circuit 255, and FIG. 4 is a block diagram showing a configuration on the synthesis side in the conventional system. In the figure, 110: buffer memory circuit, 120: subtraction circuit,
250, 350: synthesis filter circuit, 200: weighting circuit, 170:
Impulse response calculation circuit, 180: autocorrelation function calculation circuit,
210: cross-correlation function calculation circuit, 220: drive signal calculation circuit, 24
0, 340: drive signal restoration circuit, 130: pitch analysis circuit, 140: K
Parameter calculation circuit, 150: pitch coding circuit, 160: K parameter coding circuit, 225, 310: noise memory, 230: coding circuit, 255, 335: interpolation circuit, 260: multiplexer, 29
0: demultiplexer, 300: restoration circuit, 320: pitch decoding circuit, 330: K parameter decoding circuit.

Claims (3)

(57) [Claims] 1. A transmitting side receives a discrete voice signal, divides the signal into predetermined time intervals, and extracts a spectrum parameter representing a short-time spectrum envelope and a pitch parameter representing a pitch period from the voice signal. The audio signal of the time interval is divided into small sections corresponding to the pitch period, and a sound source for representing the audio signal is represented by a pulse train of a representative section of the small sections or a combination of a pulse train and noise, A combination of the information representing the sound source, the pitch parameter, and the spectrum parameter is output, and the reception side temporally generates a pulse train of the representative section based on the pitch parameter and the information representing the sound source for each pitch cycle. The audio signal is restored using the spectral parameters by restoring the driving sound source signal by performing a process of giving a smooth change. High-efficiency speech coding method, characterized by synthesis. 2. A parameter calculation circuit for dividing an input speech signal into predetermined time intervals, extracting a spectrum parameter representing a short-time spectrum envelope and a pitch parameter representing a pitch period from the speech signal, and encoding the parameter. The audio signal of the time interval is divided into small sections corresponding to the pitch period, and a sound source for representing the audio signal is obtained by obtaining a pulse train or a combination of a pulse train and noise in a representative section of the small sections. And a multiplexer circuit for combining and outputting an output code of the parameter calculation circuit and an output code of the drive signal calculation circuit. apparatus. 3. A code sequence in which a code representing a pitch parameter, a code representing a spectrum parameter, and a code representing excitation information are combined, and a code representing the pitch parameter, a code representing the spectrum parameter, and the excitation information are used as a code sequence. A demultiplexer circuit that separates and decodes a code to be represented, and gives a temporally smooth change to the pulse train of the representative section for each pitch cycle based on the decoded pitch parameter and the decoded excitation information. Processing, when a pulse train and noise are used as a sound source, a driving sound source signal restoring circuit that generates a pulse train and noise based on the sound source information to restore a driving sound source signal, and the driving sound source signal and the decoded spectrum. And a synthesis filter circuit for synthesizing and outputting a speech signal based on parameters. Apparatus.