JPH1097294A - Voice coding device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the sound quality of the sound source generating section in a CELP(code excited linear prediction) type voice coding device. SOLUTION: A pitch peak position calculator 12 determines the pitch peak position of an adaptive code vector, an amplitude emphasizing window generator 13 generates a window for emphasizing the amplitude of the pitch peak position and an amplitude emphasizing window applicator 16 emphasizes the amplitude of the noise code vector corresponding to the pitch peak position. Or it determines the search positions of pulses so that they are dense near the pitch peak position and sparse in the other regions, and, based on the determined search positions, searches pulse positions. Or it adaptively switches backward the source constitution to improve sound quality and suppress the propagation of the effects of transmission line error, making use of the pitch peak position and pitch period information in the just precedent subframe and the pitch period information in the present subframe.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a CELP in a mobile communication system for coding and transmitting a voice signal.
The present invention relates to a (Code Excited Linear Predicion) type speech encoding device.

[0002]

2. Description of the Related Art A CELP-type speech coding apparatus divides speech into a certain frame length, performs linear prediction of speech for each frame, and obtains a prediction residual (excitation signal) by linear prediction for each frame. The coding is performed using an adaptive code vector composed of a waveform and a noise code vector. The adaptive code vector and the noise code vector are, as shown in FIG. 31, the case where the adaptive code vector and the noise code vector stored in the adaptive codebook 1 and the noise codebook 2 are used as they are, respectively, as shown in FIG. In some cases, a noise code vector obtained by synchronizing the adaptive code vector from the adaptive codebook 1 and the noise code vector from the noise codebook 2 with the pitch synchronization L of the adaptive codebook 1 is used. FIG.
1 is a configuration of a noise excitation vector generation unit in a CELP-type speech coding apparatus disclosed in Japanese Patent Application Laid-Open No. 19795 and Japanese Patent Application Laid-Open No. 5-19796. In FIG. 32,
An adaptive code vector is selected from the adaptive codebook 1, and a pitch period L is output. The noise code vector selected from the noise codebook 2 is periodicized by the periodicizer 3 using the pitch period L. Periodization is performed by cutting out the noise code vector by the pitch period from the beginning and connecting it repeatedly a plurality of times until it reaches the subframe length.

[0003]

However, in the above-described conventional CELP speech coding apparatus for pitch-performing a noise code vector, the pitch period component remaining after removing the adaptive code vector component is determined by using the noise code vector as a pitch. Since it is removed by cycling at intervals, the phase information existing in one pitch waveform, that is, information on where the peak of the pitch pulse exists, is not actively used, and the sound quality is improved. There was a limit.

An object of the present invention is to solve such a conventional problem, and an object of the present invention is to provide a speech encoding device capable of further improving speech quality.

[0005]

To achieve the above object, the present invention enhances the amplitude of a noise code vector corresponding to the pitch peak position of an adaptive code vector by increasing the phase present in one pitch waveform. The information is used to improve the sound quality.

Further, the present invention uses a noise code vector limited only to the vicinity of the pitch peak of an adaptive code vector to reduce sound quality deterioration even when the number of bits allocated to the noise code vector is small. is there.

The present invention also limits the search range of the pulse position using the position of the pitch peak and the pitch period of the adaptive code vector, thereby reducing sound quality degradation even when the number of bits representing the pulse position is small. While narrowing the search range.

In the present invention, when the search range of the pulse position is limited by using the position of the pitch peak and the pitch period of the adaptive code vector, particularly, the pulse position search accuracy of the 1 to 2 pitch waveform is made fine, This is to improve the sound quality of a voiced portion of a voice having a short pitch cycle.

The present invention is also intended to improve the sound quality by changing the number of pulses of the pulse sound source according to the value of the pitch period.

In the present invention, the sound quality is improved by performing a pulse sound source search after determining the pulse amplitudes in the vicinity of the pitch peak position of the adaptive code vector and other portions in advance.

The present invention also utilizes the first-stage quantized information of the pitch gain as mode information for switching the noise codebook by performing the multi-stage quantization of the pitch gain and performing the first-stage quantized information immediately after searching for the adaptive codebook. The encoding efficiency is improved as much as possible.

The present invention also improves the voice quality by controlling the switching of the search position of the pulse sound source using the quantization pitch period information or the quantization pitch gain information in the immediately preceding subframe or the current subframe. It is designed to work.

According to the present invention, the amount of information to be transmitted is determined by determining the continuity of the phase between subframes in the backward direction and applying the phase adaptation process only to the subframes determined to have the continuous phase. The switching of the phase adaptive processing is performed without increasing the sound quality, thereby improving the sound quality. If the fixed codebook is used when the phase adaptation process is not performed, it is possible to obtain the effect of preventing transmission path error propagation.

The present invention also determines whether or not to apply the phase adaptation processing based on the concentration of signal power in the vicinity of a pitch peak position in an adaptive code vector. The adaptive processing is switched to improve the voice quality. If the fixed codebook is used when the phase adaptation process is not performed, it is possible to obtain the effect of preventing transmission path error propagation.

The present invention also provides a CELP-type speech coding apparatus for searching for an excitation pulse at a relative position from a pitch peak position, by indexing pulse positions sequentially from the head of a subframe. This is to prevent the effect of a transmission path error occurring in a frame from propagating to a subsequent frame without a transmission path error.

The present invention also provides a CELP-type speech coding apparatus for searching for an excitation pulse at a relative position from a pitch peak position, in which a pulse position is indexed sequentially from the head of a subframe and different pulses of the same index. Also, by assigning pulse numbers sequentially from the beginning of the subframe, the effect of a transmission line error generated in a certain frame is prevented from propagating to a subsequent frame without a transmission line error. is there.

The present invention also provides a CELP-type speech coding apparatus that performs a sound source pulse search at a relative position from a pitch peak position, wherein not all the pulse search positions are represented by relative positions, but only a part near the pitch peak. Is expressed as a relative position, and the remaining portion is set at a predetermined fixed position, so that the effect of a transmission line error generated in a certain frame is prevented from propagating to a subsequent frame without a transmission line error. Things.

According to the present invention, when the pitch peak position is obtained, the search for the pitch peak position is not performed for the entire target signal, but for the search for the pitch peak position in the extracted signal having the pitch period length. Is provided so that the leading pitch peak position can be more accurately extracted.

In the present invention, in a portion where the pitch period is continuous between subframes, that is, a portion considered to be a voiced stationary portion, the pitch peak position in the immediately preceding subframe, the pitch period in the immediately preceding subframe and the current By predicting the pitch peak position in the current sub-frame using the pitch period in the sub-frame, and limiting the range of the pitch peak position in the current sub-frame based on the predicted pitch peak position, in the voiced stationary part The pitch peak position can be extracted so that phase discontinuity does not occur.

The present invention also has a subframe length of 10 ms.
And when the amount of information allocated to the noise codebook information is relatively small, such as about 15 bits per subframe, and when applying a pulse excitation as a noise codebook, the number of pulses is reduced and each pulse is reduced. , And a mode in which the number of pulses is increased instead of coarsening the position information of each pulse. By improving the quality of the portion and increasing the number of pulses, it is possible to suppress the deterioration of the voice quality due to coarse position information of each pulse.

[0021]

1 is a block diagram showing a configuration of a CELP type speech coding apparatus according to a first embodiment of the present invention; FIG. 2 is a block diagram showing a configuration of a CELP type speech coding apparatus according to a first embodiment of the present invention; The present invention provides a sound encoding device that can improve sound quality by using phase information existing in one pitch waveform.

According to a second aspect of the present invention, in the speech encoding apparatus according to the first aspect, the speech generating unit multiplies the noise code vector by an amplitude emphasis window synchronized with a pitch period of the adaptive code vector. Thus, the amplitude of the noise code vector corresponding to the position of the pitch peak of the adaptive code vector is emphasized, and the sound quality can be improved by emphasizing the amplitude of the noise excitation vector in synchronization with the pitch cycle.

According to a third aspect of the present invention, in the speech encoding apparatus according to the second aspect, the speech generation unit uses a triangular window centered on a pitch peak position of the adaptive code vector as an amplitude emphasizing window. This makes it possible to easily control the amplitude emphasis window length.

According to a fourth aspect of the present invention, a CEL
In a P-type speech coding apparatus, a speech coding apparatus including a sound source generation unit that uses a noise code vector limited only to the vicinity of a pitch peak of an adaptive code vector is provided. By using the vector, even when the number of bits allocated to the noise code vector is small, deterioration of sound quality can be reduced, and sound quality can be improved in a voiced part where residual power is concentrated near the pitch pulse.

According to a fifth aspect of the present invention, in a CELP type speech coding apparatus using a pulse excitation as a noise codebook, a search range of a pulse position is determined by a pitch period and a pitch peak position of an adaptive code vector. This is a speech encoding device including a sound source generation unit, and can reduce sound quality degradation even when the number of bits allocated to pulse positions is small.

According to a sixth aspect of the present invention, in the speech coding apparatus according to the fifth aspect, the excitation generating section is arranged so that the vicinity of the pitch peak position of the adaptive code vector is dense, and the other parts are sparse. The search range of the pulse position is determined in such a manner as to make it possible to finely search for a portion where the probability of forming a pulse is high, so that the sound can be improved.

According to a seventh aspect of the present invention, the search range of the pulse position is switched according to the pitch period.
7. The speech encoding apparatus according to claim 6, wherein the search range of the pulse position is expanded or contracted based on the pitch cycle. The quality can be improved.

According to an eighth aspect of the present invention, when a plurality of pitch peaks exist in the adaptive code vector, the search range of the pulse position is limited so that the positions of at least two pitch peaks are included in the search range. 8. The speech encoding apparatus according to claim 7, wherein the effect of a case where the position of the detected leading pitch peak is incorrect is reduced.
Further, since it is possible to cope with shape changes of the waveform near the first pitch peak and the waveform near the second pitch peak, it is possible to improve the sound quality.

The ninth aspect of the present invention provides a CELP
In the speech coding apparatus of the type, the speech coding apparatus includes a sound source generation unit that switches a noise codebook according to a result of speech analysis. Improvement can be achieved.

The invention according to claim 10 of the present invention provides a CEL
The P-type speech coding apparatus is a speech coding apparatus including a sound source generation unit that switches a random codebook using transmission parameters extracted before performing a random codebook search, and has already been determined to be transmitted. Since the random codebook is switched using the existing information, the random codebook can be switched without increasing the amount of information.

An eleventh aspect of the present invention is the audio encoding apparatus according to any one of the fifth to eighth aspects, wherein the number of pulses is switched according to an analysis result of the audio signal. Since the number of pulses is switched according to, the sound quality can be improved.

The invention according to claim 12 of the present invention has a configuration in which the number of pulses is switched using information extracted before performing a random codebook search. The speech encoding apparatus according to any of the above, wherein the number of pulses is switched using information that has already been determined to be transmitted, so that the number of pulses can be switched without increasing the amount of information to be transmitted.

According to a thirteenth aspect of the present invention, the speech codec according to any one of the fifth to eighth or eleventh or twelfth aspects, further comprising a sound source generation unit for switching the number of pulses according to a pitch period. Since the number of pulses is switched using the pitch period, the number of pulses can be switched without increasing transmission information. In addition, since the optimum number of pulses differs depending on the pitch period, the sound quality can be improved.

According to a fourteenth aspect of the present invention, there is provided the speech encoding apparatus according to the thirteenth aspect, wherein the number of pulses is switched between a case in which the pitch cycle varies between consecutive subframes and a case in which the variation is not so. Since the number of pulses used in the rising portion and the steady portion of the voiced portion of the voice signal is switched, the voice quality can be improved.

According to a fifteenth aspect of the present invention, in the noise code vector generation unit using a pulse excitation as a noise excitation, the pulse amplitude is determined prior to the pulse position search. The speech encoding apparatus according to any one of claims 11 to 14, wherein the pulse sound source has a variation in amplitude, thereby improving speech quality. Further, since the amplitude is determined before the pulse search, an optimum pulse position can be determined for the amplitude.

According to the invention of claim 16 of the present invention, in the noise code vector generation section using a pulse excitation as a noise excitation, the pulse amplitude is changed near the pitch peak of the adaptive code vector and other portions. Since it is a speech coding device, it changes the amplitude near the pitch peak of the sound source signal and the rest of the pitch, so that the pitch structure of the sound source signal can be expressed efficiently, improving speech quality and quantizing pulse amplitude information. Efficiency can be improved.

According to a seventeenth aspect of the present invention, there is provided the speech encoding apparatus according to the thirteenth aspect, wherein the number of pulses of the pulse sound source to be used is determined based on the pitch period, either statistically or by learning. Since the optimal number of pulses for the cycle is determined statistically or by another learning method, it is possible to improve speech quality.

The invention according to claim 18 of the present invention provides a CEL
The P-type speech coding apparatus includes a sound source generation unit that quantizes pitch gains in multiple stages. In the first stage, a value obtained immediately after the adaptive codebook search is used as a quantization target, and in the second and subsequent stages, all sound source searches are completed. This is a speech coding apparatus that uses a difference between a pitch gain determined in a closed loop search and a value quantized in a first stage as a quantization target, and is driven by the sum of an adaptive codebook and a fixed codebook (noise codebook). In a CELP type speech coding apparatus that generates a vector, information obtained before searching for a fixed codebook (noise codebook) is quantized and transmitted. Therefore, a fixed codebook (noise codebook) is added without adding independent mode information. ) Can be switched, and audio information can be efficiently encoded.

According to a nineteenth aspect of the present invention, in the speech coding apparatus according to the eighteenth aspect, a fixed codebook is switched by using a quantization value of a pitch gain obtained immediately after an adaptive codebook search. , Claims 9 to 12
Alternatively, the speech coding apparatus according to any one of claims 15 to 17, wherein the pitch gain obtained before searching for the fixed codebook and the pitch gain calculated after searching for the fixed codebook do not greatly differ. Thus, it is possible to switch the mode of the fixed codebook without adding mode information, and it is possible to improve speech quality.

According to a twentieth aspect of the present invention, the fixed codebook is switched based on a change in pitch period between subframes.
9. The speech coding apparatus according to any one of claims 9 to 9, wherein the speech encoding unit determines whether or not the voiced / voiced stationary unit is used by utilizing continuity between subframes of the pitch period and the like. The sound quality can be improved by switching between a sound source that is effective for the sound source and a sound source that is effective for the other parts (silent / rising portion, etc.).

According to a twenty-first aspect of the present invention, any one of the ninth to twelfth or fifteenth to seventeenth aspects switches the fixed codebook using the pitch gain quantized in the immediately preceding subframe. The voice coding apparatus according to any one of (1) to (3), by utilizing the continuity between pitch gain subframes and the like, to determine whether or not the voiced / voiced stationary section is effective. By switching between the sound source and a sound source that is effective for other parts (unvoiced, rising part, etc.), it is possible to improve the sound quality.

According to a twenty-second aspect of the present invention, the fixed codebook is switched based on a change in pitch period between subframes and a quantized pitch gain.
18. The speech coding apparatus according to claim 15, wherein a determination is made as to whether or not the voiced / voiced stationary section is to be used, using information on a pitch period and a pitch gain, which are transmission parameters. By performing the switching between the sound source effective for the voiced / voiced stationary part and the sound source effective for the other parts (unvoiced / rising part, etc.), the sound quality can be improved.

According to a twenty-third aspect of the present invention, the speech coding apparatus according to any one of the nineteenth to twenty-second aspects uses a pulse excitation codebook as the fixed codebook, and includes a pulse excitation codebook as the noise codebook. Is used, it is possible to reduce the amount of memory required for the random codebook and the amount of calculation at the time of searching for the random codebook, and it is possible to further improve the expression of the rise of the voiced part.

According to a twenty-fourth aspect of the present invention, in a CELP type speech coding apparatus for performing speech coding processing for each subframe having a predetermined time length, a phase in a current subframe and a phase in a previous subframe are compared. A speech encoding device that determines whether a phase is continuous and switches a sound source used when it is determined that the phases are continuous and when it is determined that the phases are not continuous.
The sound source configuration can be realized by separating the part from the rest.
Sound quality can be improved.

According to a twenty-fifth aspect of the present invention, the pitch peak position in the current subframe is determined by using the pitch peak position in the immediately preceding subframe, the pitch period in the immediately preceding subframe, and the pitch period in the current subframe. The position is predicted, and the phase in the immediately preceding subframe and the current position are determined depending on whether or not the pitch peak position in the current subframe obtained by this prediction is close to the pitch peak position obtained only from the data in the current subframe. The CELP-type speech encoding apparatus according to claim 24, wherein it is determined whether or not the phase in the subframe is continuous, and the encoding processing method of the excitation is switched based on the determination result. The judgment result is obtained by using the information The do not need to be transmitted using the new transmission information.

According to a twenty-sixth aspect of the present invention, when it is determined that the phase in the immediately preceding subframe and the phase in the current subframe are continuous, the phase adaptation processing is performed on the noise codebook. 26. When it is determined that the phase in the immediately preceding subframe and the phase in the current subframe are not continuous, the phase adaptation processing is not performed on the noise codebook.
It is possible to perform effective phase adaptation processing. In addition, since the continuity of the phase between subframes is determined backward, there is no need to newly transmit switching information as to whether or not to apply the phase adaptation processing. Further, when the phase adaptation processing is not applied, by using a fixed codebook, it is possible to obtain the effect of suppressing the propagation of the influence of the transmission path error.

According to a twenty-seventh aspect of the present invention, there is provided a CELP-type speech coding apparatus for performing a speech coding process for each subframe having a predetermined time length, in the vicinity of a pitch peak of an adaptive code vector in a current subframe. Is a speech encoding apparatus that switches the encoding processing method of the excitation signal based on the signal power concentration level in, and does not require new transmission information for switching the excitation configuration (the encoding processing method of the excitation signal). In addition, the sound source configuration can be adaptively switched.

[0048] The invention according to claim 28 of the present invention is arranged such that the ratio of the signal power in the vicinity of the pitch peak of the adaptive code vector in the current subframe to the entire signal of one pitch period length is equal to or more than a predetermined value. 28. The speech encoding apparatus according to claim 27, wherein the phase adaptation process is performed on the noise codebook and the phase adaptation process is not performed on the noise codebook if the value is less than a predetermined value. , The phase adaptive processing can be adaptively controlled (switched) depending on the strength of the pulse characteristic of, and the sound quality can be improved. Control (switching) of phase adaptation processing
No new transmission information is required. Furthermore, if the fixed codebook is used when the phase adaptation process is not performed, it is possible to obtain an effect of suppressing the propagation of the influence of the transmission path error.

According to a twenty-ninth aspect of the present invention, as a phase adaptation process, a pulse position search is performed densely near a pitch peak, and a sparse pulse position search is performed in a portion other than the vicinity of a pitch peak. Claim 2 applied to a sound source
29. The speech coding apparatus according to claim 26, wherein a pulse excitation is used for the noise codebook, so that the amount of memory required for the noise codebook and the amount of calculation at the time of searching for the noise codebook can be reduced. The expression of the rising edge can be improved.

According to a thirty-fifth aspect of the present invention, the indices indicating the positions of the pulses are added so as to be arranged in order from the head of the subframe. 30. The speech encoding device according to claim 23, wherein an index representing a pulse position is assigned from the beginning of the subframe such that the index number is smaller near the beginning of the subframe. This makes it possible to reduce the deviation of the pulse position that occurs when the pitch peak position is incorrect, thereby reducing the propagation of the influence of the transmission path error.

According to the invention of claim 31 of the present invention, when the index numbers are the same, the pulse numbers are sequentially numbered from the head side of the subframe, and the vicinity of the pitch peak position is dense, and the parts other than the pitch peak are densely arranged. 31. The speech coding apparatus according to claim 30, wherein a search position of each pulse is determined so as to be sparse. Since the number of each pulse is determined, the numbering method of the pulse is also defined in addition to the index of the pulse. Can be further reduced.

In the invention according to claim 32 of the present invention, a part of the pulse search position is determined by the pitch peak position, and the other pulse search positions are predetermined fixed positions irrespective of the pitch peak position. The speech encoding apparatus according to any one of claims 5 to 8, 11 to 17, or 23 or 29, wherein the position of the excitation pulse is incorrect even when the pitch peak position is incorrect. Since the probability decreases, propagation of the influence of the transmission path error can be suppressed.

According to a thirty-third aspect of the present invention, when determining the pitch peak position of a voice or sound source signal having a predetermined time length, only the pitch period length is cut out from the signal and the pitch peak is included in the cut out signal. The speech code according to any one of claims 1 to 8, or 11 to 17, or 19 to 23, or 25 to 32, comprising a pitch peak position calculation means for determining a position. In order to select a pitch peak from a one-pitch waveform, it is sufficient to simply search for a point where the amplitude value (absolute value) becomes maximum, and a subframe includes a waveform that exceeds one pitch period. However, the pitch peak position can be accurately obtained.

According to the thirty-fourth aspect of the present invention, when only the pitch period length is cut out from the signal, the pitch peak position is determined by using the entire signal without cutting out one cycle length, and the pitch peak position is determined. 34. One pitch period length is cut out using the cut-out pitch peak position as a cut-out start point, and the pitch peak position is determined in the cut-out signal.
It is possible to avoid a phenomenon in which a second peak in one pitch waveform is set as a pitch peak position, which occurs when a pitch peak position is determined using the entire signal. That is, it is possible to avoid a pitch peak position extraction error caused by the synchronization between the pitch period and the subframe length.

According to a thirty-fifth aspect of the present invention, there is provided a CELP-type speech coding apparatus for performing speech coding processing for each subframe having a predetermined time length when calculating a pitch peak position in a current subframe. If the difference between the pitch cycle in the immediately preceding subframe and the pitch cycle in the current subframe is within a predetermined range, the pitch peak position in the immediately preceding subframe, and the pitch cycle in the immediately preceding subframe, Predict the pitch peak position in the current subframe using the pitch period in the current subframe, and use the pitch peak position in the current subframe obtained by this prediction to determine the range of the pitch peak position in the current subframe Is limited in advance and a pitch peak position search is performed within that range. The speech encoding apparatus according to any one of claims 1 to 8 or claim 11 to claim 17 or claim 19 to claim 23 or claim 25 to claim 32, wherein the pitch of the immediately preceding subframe is In order to determine the pitch peak position of the current subframe in consideration of the peak position, if the pitch peak position is obtained only from the current subframe, the second peak position in one pitch peak waveform may be erroneously detected. This is a technique for avoiding erroneous detection.

According to a thirty-sixth aspect of the present invention, in a CELP type speech coding apparatus for performing speech coding processing for each subframe having a predetermined time length, a pulse excitation is used as a noise codebook and a noise codebook is used. Mode of at least 2
The number of sound source pulses can be changed by switching the mode. At least one is a mode in which the position information of each pulse is sufficiently small and the number of pulses is small, and the other is the position of each pulse. This is a speech encoding device that lacks information but has a large number of pulses, and is a speech encoding device that performs mode switching by transmitting mode switching information. The quality of the voiced rising portion of the signal can be improved, and a mode having a large number of sound source pulses with insufficient position information can be effectively used.

According to the thirty-seventh aspect of the present invention, when the pitch period is short, the search range of the source pulse is limited to a narrow range corresponding to the pitch period, thereby reducing the position information of the source pulse. 37. The speech encoding apparatus according to claim 36, wherein the number of excitation pulses is increased by increasing the number of excitation pulses, and position information of the excitation pulses per pitch period is sufficiently maintained for an excitation signal having a short pitch period. The number of sound source pulses can be increased as it is, and the sound quality can be improved.

According to the thirty-eighth aspect of the present invention, in the mode in which the position information of each pulse is insufficient but the number of pulses is large, the search position of the sound source pulse is densely located near the pitch peak position, and in other parts. The speech encoding apparatus according to claim 36 or 37, wherein a search range of the pulse position is determined so that the search position of the excitation pulse is sparse. Since the position information is concentrated, the use efficiency of a mode having a large number of sound source pulses where the position information of the sound source pulse is insufficient can be improved.

The invention according to claim 39 of the present invention provides the CELP according to claim 36 or 37 or 38.
In a speech coding apparatus of the type, in a sound source mode in which the number of pulses is small and the position information is sufficient, a part of the position information is assigned to an index representing a noisy excitation code vector, and a new It is possible to cope with an unvoiced consonant part or a noise-like input signal without providing a mode.

According to a forty-third aspect of the present invention, there is provided a computer-readable recording medium having recorded thereon a program for executing the function of the speech coding apparatus according to any one of the first to thirty-ninth aspects. By reading such a recording medium by a computer, the function of the audio encoding device can be realized.

Hereinafter, the excitation generator of the speech encoding apparatus according to the embodiment of the present invention will be described with reference to FIGS.

(Embodiment 1) FIG. 1 shows a first embodiment of the present invention, in which a sound source generator of a speech coding apparatus for enhancing the amplitude of a noise code vector corresponding to a pitch peak position of an adaptive code vector. Is shown. In FIG. 1, reference numeral 11 denotes an adaptive codebook that outputs an adaptive code vector to a pitch peak position detector 12. Pitch position calculator to output to
Reference numeral 3 denotes an amplitude emphasis window generator which receives the pitch peak position output from the pitch peak position calculator 12 and outputs an amplitude emphasis window to the amplitude emphasis window multiplier 16, and 14 stores a noise code vector and a periodizer. 15, a random codebook to be output to
Reference numeral 15 denotes a periodicizer which receives the noise code vector output from the noise codebook 14 and the pitch period L as input, converts the noise code vector to a pitch period, and outputs the pitch period to an amplitude emphasizing window multiplier 16. Is an amplitude emphasizing window multiplier that receives the amplitude emphasizing window output from the controller and the noise code vector output from the periodizer 15 as inputs, multiplies the noise code vector by the amplitude emphasizing window, and outputs a final noise code vector. .

The operation of the excitation generating section of the CELP speech coding apparatus configured as described above will be described with reference to FIG. The pitch peak position calculator 12 determines the position of the pitch pulse existing in the adaptive code vector using the input adaptive code vector. The position of the pitch pulse is
This can be performed by maximizing the normalized cross-correlation between the impulse train arranged in the pitch period and the adaptive code vector. It is also possible to minimize the error caused by passing the impulse train arranged in the pitch cycle through the synthesis filter and by minimizing the error caused by passing the adaptive code vector through the synthesis filter.

The amplitude emphasis window generator 13 generates an amplitude emphasis window based on the pitch pulse position determined by the pitch peak position calculator 12. Various types of amplitude emphasis windows can be used. For example, a triangular window centered on the pitch pulse position is advantageous in that the window length can be easily controlled.

FIG. 2 shows the correspondence between the shape of the amplitude enhancement window output from the amplitude enhancement window generator 13 and the shape of the adaptive code vector. The position indicated by the broken line in the figure is the pitch pulse position determined by the pitch peak position calculator 12.

The periodizer 15 pitch-cycles the random code vector output from the random codebook 14. Pitch periodicization is a method of periodicizing a random code vector with a pitch period. The stored vector of the random codebook is cut out from the beginning by the pitch period L, and connected repeatedly a plurality of times until the subframe length is reached. Done by However, the pitch period is set only when the pitch period is equal to or less than the subframe length.

The amplitude emphasizing window generator 16 adds the noise emphasizing window generator 13 to the noise code vector output from the periodizer 15.
Multiply by the amplitude emphasis window output from.

As described above, according to the first embodiment, the sound quality can be improved by utilizing the phase information existing in one pitch waveform.

In FIG. 1, the excitation part of the CELP-type speech coding apparatus for performing the periodicization of the random code vector has been described. However, the random code vector stored in the random code book as shown in FIG. 11 is used as it is. Common CEL
The present invention is also applicable to the sound source portion of the P-type speech coding apparatus, and an example is shown in FIG. In FIG. 3, 21 is an adaptive codebook, 22 is a peach peak position calculator, 23 is an amplitude emphasis window generator, 24 is a noise codebook, and 25 is an amplitude emphasis window multiplier, which does not synchronize a noise source with pitch synchronization. This is different from the sound source generation unit in FIG.

(Embodiment 2) FIG. 4 shows a second embodiment of the present invention, which has a configuration in which a sound source combining a pulse train sound source and a noise sound source is applied to the rising portion of a voiced portion of a voice signal. FIG. 3 shows a sound source generation unit of a speech coding device that emphasizes the amplitude of a noise code vector corresponding to a pulse position of a pulse train excitation with respect to a CELP type speech coding device. In FIG. 4, reference numeral 31 denotes an amplitude enhancement window generator 32.
And a pulse train source composed of impulse trains output to the adder 33 and arranged at intervals of the pitch period L placed at the position of the pitch pulse. 32 emphasizes the noise code vector amplitude at a position corresponding to the pulse position of the pulse train. An amplitude emphasis window generator for generating an amplitude emphasis window for output to the multiplier 35 and adding the pulse train excitation and the noise code vector after the amplitude emphasis window output from the multiplier 35,
An adder that outputs as an excitation vector; 34 is a noise source represented by a noise code vector and output to a multiplier 35
A multiplier 35 multiplies the noise source vector output from the noise source 34 by the amplitude enhancement window output from the amplitude enhancement window generator 32.

The operation of the sound source generator configured as described above will be described with reference to FIG. Pulse train sound source 3
1 is a pulse train whose pulse position and interval are determined by the pitch period L and the initial phase P.
And the initial phase P is separately calculated outside the sound source generation unit. The pulse train sound source may be one in which impulses are arranged, but the performance is better if the impulse existing between the sampling points can be expressed. Similarly, the performance of the initial phase (the position of the first pulse) is better if it is expressed with fractional precision that can represent between sampling points. However, if the number of bits that can be allocated to this information is not enough, Good performance can be obtained even with integer precision, and search for position determination is easy.

The amplitude emphasis window generator 32 is a window for emphasizing the amplitude of the noise source vector at the position corresponding to the pulse position of the pulse train excitation vector, and includes the amplitude emphasis window described in the first embodiment. It is similar. A triangular window centered on the position of the pulse can be used.

The adder 33 outputs the pulse train excitation vector 31
And the noise source vector 34 multiplied by the amplitude emphasizing window by the multiplier 35 and output as an excitation source vector.

Although not shown in FIG. 4, if the pulse train excitation vector and the noise excitation vector are each multiplied by an appropriate gain before being input to the adder 33, the sound source generation with higher expressiveness can be achieved. Department. However, in that case, it is necessary to separately transmit the gain information. Also,
When fixing the gain of the pulse train excitation vector and the noise excitation vector, adjust the power of the pulse train excitation vector and the power of the noise excitation vector to be equal so that the pulse train excitation vector is not buried in the noise excitation vector. Gain adjustment is required.

As described above, according to the second embodiment, the sound quality can be improved by emphasizing the amplitude of the noise source vector in synchronization with the pitch period.

(Embodiment 3) FIG. 5 shows a third embodiment of the present invention.
5 shows a sound source generation unit of a speech coding apparatus using a noise code vector limited to only the vicinity of a pitch peak of an adaptive code vector.

In FIG. 5, reference numeral 41 denotes an adaptive codebook for outputting an adaptive code vector, and reference numeral 42 denotes an input of the adaptive code vector output from the adaptive codebook 41 and the pitch period L, and the position (phase information) of a pitch peak is determined by noise. The phase searcher 43 outputs to the code vector generator 44 a noise code vector whose vector length is limited only in the vicinity of the pitch peak, and outputs the noise code vector near the pitch pulse position to the noise code vector generator 44. A noise codebook 44 limited to near the pitch pulse position is input with the noise code vector output from the noise codebook 43 limited to the near pitch pulse position, the phase information output from the phase searcher 42, and the pitch period L as inputs. A random code vector generator that outputs the vector to the periodicizer 45, and 45 is a signal output from the random code vector generator 44. Noise code vector and the pitch cycle L
, And outputs a final random code vector.

The operation of the sound source generation unit of the speech coding apparatus configured as described above will be described with reference to FIG. The phase searcher 42 uses the adaptive code vector output from the adaptive codebook 41 to determine the position (phase) of the pitch pulse existing in the adaptive code vector. The position of the pitch pulse can be determined by maximizing the normalized cross-correlation between the impulse train arranged in the pitch cycle and the adaptive code vector. Further, it is possible to obtain the impulse trains arranged in the pitch cycle through a synthesis filter, and to minimize the error caused by passing the adaptive code vector through the synthesis filter, thereby obtaining the error more accurately.

Noise code book 4 limited to near pitch pulse position
Numeral 3 stores a noise code vector to be applied to the vicinity of the pitch peak of the adaptive code vector, and the vector length is fixed regardless of the pitch period and the frame (subframe) length. The range near the pitch peak may be equal in length before and after the pitch peak,
When the range after the pitch peak is longer than before, the sound quality is less deteriorated. For example, when the vicinity range is set to 5 msec, it is better to set 0.625 msec before the pitch peak and 4.375 msec after the pitch peak, rather than 2.5 msec before and after the pitch peak. As the vector length, the subframe length is 10
In the case of msec, if the vector length is about 5 msec, almost the same sound quality as when the vector length is 10 msec or more can be realized.

The noise code vector generator 44 arranges the noise code vector output from the pitch pulse position limited noise codebook 43 at the position of the pitch pulse determined by the phase searcher 42.

FIGS. 6 and 7 illustrate a method of arranging the noise code vector output from the pitch pulse position limited type noise codebook 43 at a position corresponding to the pitch pulse position by the noise code vector generator 44. FIG. is there. Basically, as shown in FIG. 6A, a pitch pulse position limited noise code vector is arranged near the pitch pulse position. In FIG. 6, a portion (hatched portion) indicated as a pitch period range is a portion which is a target when the pitch period is set by the period unit 45. In the case of FIG. 6A, it is not necessary to perform the pitch period in the noise code vector generator 44. However, in the case of FIG. 6B, the position of the pitch pulse is near the subframe boundary. , The first half (the part before the subframe boundary) of the noise code vector output from the pitch pulse position limited type noise codebook 43 cannot be periodicized by the periodicizing unit 45 (the periodicizing unit 45). In, the vector cut out from the subframe boundary by the pitch period length is repeatedly arranged in the pitch period.) The noise code vector generator 44 is operated to make the pitch period in advance.
In addition, when the pitch pulse position is located immediately before the subframe boundary, if the pitch pulse is cut out from the head of the subframe and is periodicized, the latter half of the limited vector near the pitch pulse position is not properly pitched.
As shown in (a), the noise vector generator 44 operates to make the pitch period in the negative direction of the time axis. However, if the pitch pulse position does not exist between the head of the subframe and the pitch period length, this periodization is not necessary. As described above, by performing the pitch period prior to the pitch period unit 45, the pitch period unit 45 effectively uses all the parts of the pitch position vicinity limited vector so that the pitch period unit 45 performs the pitch period. I have. When the pitch cycle is shorter than the vector length limited to the vicinity of the pitch pulse position, the pitch cycle is performed by cutting out the limited vector by the pitch cycle length. In this case, although there are various ways of cutting out, the cutting is performed so that the pitch pulse position is included in the cut out vector. For example, the clipping start point is determined using the pitch pulse position and the pitch cycle, such that one pitch cycle is cut from a point one quarter pitch cycle before the pitch pulse position.

FIG. 7B shows an example of a method for extracting a noise code vector when the pitch period is shorter than the limited vector length. In this case, the pitch period length is cut out from the head of the limited noise code vector near the pitch pulse position. This eliminates the need to calculate the cutout start point every time. That is, as described above, when one pitch period is cut out from a point one quarter pitch period before the pitch pulse position, the pitch period is a variable.
Although it is necessary to calculate the one-half pitch period every time, this calculation is unnecessary because the head position of the limited noise code vector near the pitch pulse position is a fixed value. However, if the vector extracted by the pitch period length from the head of the pitch pulse position vicinity limited noise code vector does not include the portion corresponding to the pitch pulse position, the extraction is performed so that the portion corresponding to the pitch pulse position is included. It is necessary to shift the starting position.

The periodizer 45 pitch-cycles the noise code vector output from the noise code vector generator 44. Pitch periodization is a method of periodizing a noise code vector with a pitch period, and is performed by cutting out the noise code vector by the pitch period L from the head and repeating and connecting it a plurality of times until the subframe length is reached. . However, the pitch period is set only when the pitch period is equal to or less than the subframe length. In the case of a pitch cycle with fractional precision, vectors obtained by calculating points with fractional precision by interpolation are connected.

As described above, according to the third embodiment, by using the noise code vector limited only to the vicinity of the pitch peak of the adaptive code vector, even when the number of bits allocated to the noise code vector is small, Sound quality deterioration can be reduced, and sound quality can be improved in voiced parts where residual power is concentrated near the pitch pulse.

(Embodiment 4) FIG. 8 shows a fourth embodiment of the present invention, in which sound source generation of a speech coding apparatus for determining a search range of a pulse position by a pitch period and a pitch peak position of an adaptive code vector. Indicates a part. In FIG. 8, reference numeral 51 denotes an adaptive codebook which stores the past excitation source vector and outputs the selected adaptive code vector to the pitch peak position calculator 52 and the pitch gain multiplier 55.
Calculates a pitch peak position using the adaptive code vector output from the adaptive code book 51 and the pitch period L as inputs, and outputs a pitch peak position calculator to the search range calculator 53;
Reference numeral 53 denotes a search range calculator for calculating a range in which to search for a pulse sound source using the pitch peak position and the pitch period L output from the pitch peak position calculator 52 as inputs, and 54 a search range calculator. A pulse excitation searcher that searches for a pulse excitation by using the search range output from the multiplier 53 and the pitch period L as inputs, and outputs a pulse excitation vector to a pulse excitation gain multiplier 56. 55 is an adaptive excitation output from the adaptive codebook. A multiplier that multiplies the code vector by the pitch gain and outputs the result to the adder 57, 56 is a multiplier that multiplies the pulse excitation vector output from the pulse excitation searcher by the pulse excitation gain and outputs the result to the adder 57, 57 is a multiplier The output from 55 and the output from multiplier 56 are input,
This is an adder that adds and outputs as an excitation sound source vector.

The operation of the sound source generator configured as described above will be described with reference to FIG. In FIG. 8, adaptive codebook 51 cuts out an adaptive code vector by a subframe length from a point that has been traced back by a pitch period L calculated in advance outside the excitation generation unit, and outputs it as an adaptive code vector. If the pitch period L is less than the subframe length, a vector obtained by repeatedly connecting the extracted pitch period L until the subframe length is reached is output as an adaptive code vector.

The pitch peak position calculator 52 determines the position of the pitch pulse existing in the adaptive code vector using the adaptive code vector output from the adaptive code book 51. The position of the pitch pulse can be determined by maximizing the normalized cross-correlation between the impulse train arranged in the pitch cycle and the adaptive code vector. In addition, an impulse train arranged in a pitch cycle is passed through a synthesis filter,
By minimizing the error caused by passing the adaptive code vector through the synthesis filter, it is possible to obtain the error more accurately.

The search range calculator 53 uses the input pitch peak position and pitch period L to calculate the range in which to search for a pulsed sound source. That is, a perceptually important range in one pitch waveform is calculated from the pitch peak position information, and the range is determined as a search range. The specific search range determined by the search range calculator 53 is shown in FIG.
And FIG. FIG. 9A shows a case where a range of 32 samples is determined as a search range starting from a position 5 samples before the pitch peak position.
In a voiced part, if an impulse train arranged in advance at a pitch cycle is used as a pulse sound source, a pulse can be raised at the same position in the search range of the second pulse, and the sound source can be expressed efficiently. FIG. 9B shows that the pitch period is
FIG. 6 shows an example of a search range that is determined when the length is longer than that of FIG. When the pitch period is long, FIG.
When the vicinity of the pitch pulse is intensively searched as in (a), the relative search range for one pitch waveform is narrowed, and the expressible frequency band is narrowed. May deteriorate.
In such a case, as shown in FIG. 9B, instead of expanding the search range according to the pitch cycle, every other or every second sample point is searched without searching all the sample points. By providing the portion, it is possible to prevent the expressiveness of the frequency component of a specific band from being deteriorated without increasing the number of search positions.

FIG. 10 shows a method for limiting the pulse position search range densely in the vicinity of the pitch pulse position and sparsely in other portions. This limiting method is based on a statistical result in which positions where the pulse is likely to be raised are concentrated near the pitch pulse. If the pulse position search range is not limited, the probability that a pulse will be raised near the pitch pulse in a voiced part will be higher than the probability that it will be raised in other parts. However, the probability that a pulse is generated in other portions is not so small as to be negligible.
The pulse position search range limiting method shown in FIG.
In the method shown in (b), it can be said that the search range is limited based on the probability distribution of the pulse. In addition,
In FIG. 9A, when the pitch period is short and the search range of the first pulse overlaps the search range of the second pulse, the first pulse is searched so as not to overlap the search range of the second pulse. There is a method of increasing the number of pulses instead of narrowing the search range, and a method of determining a search range overlapping the search range of the second pulse (the same search range determination method as in FIG. 9A).

The pulse position searcher 54 sets a pulse sound source in the search range (position) determined by the search range calculator 53 and outputs a position where the synthesized voice is closest to the input voice. In particular, in the voiced stationary part, the subframe length is a length including a plurality of pitch pulses, and in the voiced stationary part, the first pulse position of the impulse train is determined from the search range using the impulse train arranged at the pitch period interval as the pulse sound source. It is efficient to decide. There are various ways of setting up the pulse.
When standing somewhere in the location, all combinations (8 × 8 × 8 × 8 ways) are made so that 32 locations are divided into four and one pulse is determined as one of eight assigned locations. ), And a method of searching for all combinations that select four locations from 32 locations.
In addition to the combination of the impulses having the amplitude of 1, a combination of a plurality of pulses, for example, two pulses, or a combination of impulses having different amplitudes may be used.

The gain to be multiplied by the multipliers 55 and 56 is determined by performing speech synthesis using the adaptive code vector output from the adaptive codebook 51 and the pulse excitation vector output from the pulse position searcher 54, and Is a value determined for each vector such that the error of Here, assuming that the gain multiplied by the adaptive code vector is a pitch gain and the gain multiplied by the pulse excitation vector is a pulse excitation gain, the multiplier 55 multiplies the adaptive code vector by the pitch gain and outputs the result to the adder 57. Multiplier 56
Multiplies the pulse source vector by the pulse source gain,
Output to the adder 57.

The adder 57 adds the adaptive code vector after the optimal gain multiplication output from the multiplier 55 and the pulse excitation vector after the optimal gain multiplication output from the multiplier 56 to generate an excitation excitation vector. Output.

As described above, according to the fourth embodiment, even when the number of bits allocated to the pulse is small, deterioration in sound quality can be reduced.

(Embodiment 5) FIG. 11A shows a fifth embodiment of the present invention, in which a search position for a pulse position is determined by a pitch period and a pitch peak position of an adaptive code vector. FIG. 10 shows a pulse search position determination unit, and shows the search range calculator 53 in FIG. 8 in more detail. In FIG. 11A, reference numeral 61 denotes a pulse search position pattern selector which outputs a pulse search position pattern to a pulse search position determiner 62 with a pitch period L as an input, and 62 denotes a pulse search position pattern selector. The pulse search position pattern is inputted from the pulse search position pattern 61 and the pitch peak position from the pitch peak position calculator 52.
4 is a pulse search position determiner.

The operation of the search range calculator 53 of the sound source generator configured as described above will be described with reference to FIG.
This will be described with reference to (b) and (c). The pulse search position pattern selector 61 has a plurality of types of pulse search position patterns in advance (the pulse search position pattern is composed of a set of sample point positions for performing pulse search, and is a relative position where the pitch peak position is 0). Using the pitch period L obtained by pitch analysis, which pulse search position pattern is to be used is determined, and the pulse search position pattern is output to the pulse search position determiner 62. I do.

FIGS. 11B and 11C show an example of a pulse search position pattern that the pulse search position pattern selector 61 has in advance. The scale in the figure indicates the position of the sample point, and the sample point with an arrow is the pulse search position (the part without the arrow is not searched). The scale value is a numerical value representing a relative position with the pitch peak position obtained from the adaptive code vector being 0. FIGS. 11B and 11C show the case of 80 samples per subframe. FIG.
FIG. 11B shows a search position pattern when the pitch period L is long (for example, 45 samples or more).
(C) shows a search position pattern when the pitch period L is short (for example, less than 44 samples). When the pitch period L is short, the search of the entire subframe is not performed. However, by performing the pitch period process, it is possible to raise a pulse in the entire subframe. Pitch periodicization can be easily performed by using the following equation (1) (ITU-T STUDY GROUP15-CONTRIBU
TION 152, "G.729-CODING OF SPEECH AT 8 KBIT / S USIN
G CONJUGATE-STRUCTURE ALGEBRAIC-CODE-EXCITED LINEA
R-PREDICTION (CS-ACELP) ", COM 15-152-E July 199
Five). code (i) = code (i) + β × code (i-L). . . (1) In equation (1), code () represents a pulse excitation vector, and i is a sample number (in the example of FIG. 11, 0 to 7).
9). Β is a gain value indicating the strength of the period,
When the periodicity is strong, the value is large, and when the periodicity is weak, the value is small (generally, a value of 0 to 1.0 is used). FIG.
In (c), a pulse search is performed in the range of (-4) to 48 samples (range of 53 samples). For this reason,
When the pitch period L is less than 53 (or 54), FIG.
It is also possible to use the search range pattern of (c).
However, by setting the pitch period L to be less than about 45 samples, two pitch peak positions can be included in the search range, and the pitch pulse waveform of the first cycle and 2
When the pitch pulse waveform in the cycle changes, or when the calculated pitch peak position is one
It is possible to cope with a case where the position is erroneously detected as the position before the cycle.

The pulse search position determiner 62 determines the pulse search position in the current subframe using the pulse search position pattern output from the pulse search position pattern selector, and outputs the pulse search position to the pulse position searcher 54. Since the pulse search position pattern output from the pulse search position pattern selector 62 is expressed as a relative position where the pitch peak position is 0, it cannot be used for pulse search as it is. For this reason, the sub-frame is converted into an absolute position where the head of the sub-frame is 0 and output to the pulse position searcher 54.

(Embodiment 6) FIG. 12 shows a sixth embodiment of the present invention, in which a search position of a pulse position is determined by a pitch period and a pitch peak position of an adaptive code vector and used for a pulse sound source. 3 shows a sound source generation unit of a speech encoding device having a configuration for switching the number of pulses. In FIG. 12, reference numeral 71 denotes an adaptive code vector which outputs an adaptive code vector to a pitch peak position calculator 72 and a multiplier 76. The adaptive code book 72 includes an adaptive code book and a pitch period L externally obtained by pitch analysis or adaptive code book search. The adaptive code vector output from the codebook is input, and the pitch peak position is output to a search position calculator 74. The pitch peak position calculator 73 has a pitch period L externally obtained by pitch analysis or adaptive codebook search. , And outputs the number of pulses to the search position calculator 74. The number of pulses 74 is the pitch period L obtained externally by pitch analysis or adaptive codebook search and the number of pulses output from the number of pulses determiner 73. And the pitch peak position output from the pitch peak position calculator 72 as an input, and a pulse search position is determined as a pulse position search. A search position calculator 75 outputs a pulse period L externally obtained by pitch analysis or adaptive codebook search and a pulse search position output from the search position calculator 74, and outputs a pulse used as a pulse source. A pulse position searcher 76 that determines a combination of standing positions and outputs a pulse excitation vector generated by the combination to a multiplier 77. The pulse position searcher 76 receives the adaptive code vector output from the adaptive codebook as an input, The multiplier 77 receives the pulse excitation vector output from the pulse position searcher as an input, multiplies the pulse excitation vector gain and outputs the result to the adder 78, and 78 is a multiplier. Inputs the vectors output from 76 and 77, performs vector addition, and outputs as a sound source vector It is an adder.

The operation of the excitation generating section of the CELP-type speech coding apparatus configured as described above will be described with reference to FIG. The adaptive code vector output from adaptive codebook 71 is output to multiplier 76, multiplied by the adaptive code vector gain, and output to adder 78. The pitch peak position calculator 72 detects a pitch peak from the adaptive code vector and outputs the position to the search position calculator 74.
The detection (calculation) of the pitch peak position can be performed by maximizing the inner product of the impulse train vector arranged in the pitch period L and the adaptive code vector. In addition, by maximizing the inner product of the vector obtained by convolving the impulse response of the synthesis filter with the impulse train vector arranged in the pitch period L and the vector obtained by convolving the impulse response of the synthesis filter with the adaptive code vector, the pitch can be more accurately determined. It is also possible to detect a peak position.

The number-of-pulses determiner 73 determines the number of pulses used for the pulse sound source based on the value of the pitch period L, and outputs it to the search position calculator 74. The relationship between the number of pulses and the pitch cycle is predetermined by learning or statistically. For example, when the pitch cycle is 45 samples or less, 5 is used.
Book, 4 if more than 45 samples and less than 80 samples
The number of each pulse is determined by the range of the pitch period value, such as three, for 80 samples or more. When the pitch period is short, the pulse search range can be limited to 1 to 2 pitch periods by using the pitch period processing, so that the number of pulses can be increased instead of reducing the position information. In terms of waveforms, a female voice having a short pitch cycle and a male voice having a long pitch cycle have different waveform characteristics, and the number of pulses suitable for each exists. In general, the pulse position tends to be more important than the number of pulses because a male voice has a stronger pulse, and it is better to increase the number of pulses and avoid power concentration in a female voice because the pulse is weak. Tend. From these facts, it is effective to reduce the number of pulses when the pitch period is long, and to increase the number of pulses to some extent when the pitch period is short. further,
When the number of pulses is determined in consideration of changes in the number of pulses between successive subframes, changes in the pitch period L, and the like,
It is possible to reduce the discontinuity between consecutive subframes and improve the quality of the rising part of the voiced part. In particular,
When the number of pulses determined from the pitch period L is reduced from 5 to 3 in consecutive subframes, instead of providing a hysteresis to the reduction in the number of pulses and reducing the number of pulses from 5 to 3 suddenly, By making the number of pulses four, it is possible to avoid a large change in the number of pulses between subframes.
Alternatively, when the pitch period L is greatly different between consecutive subframes, there is a high possibility that a voiced part is rising, so that the voice quality is improved by reducing the number of pulses and improving the accuracy of the pulse position. When the pitch period L of the previous sub-frame is greatly different from the pitch period L of the current sub-frame, the number of pulses is determined by a method of setting the number of pulses to 3 regardless of the value of the pitch period L of the current sub-frame. By doing so, it is possible to further improve the voice quality. When these methods are used, the influence of double pitch errors and half pitch errors in pitch analysis becomes liable to occur. Therefore, a pulse number determination method (for example, half pitch or double pitch To determine the continuity of the pitch cycle taking into account the characteristics of the pitch), or to increase the accuracy of pitch analysis as much as possible.
More effective.

The search position calculator 74 determines the position where the pulse search is to be performed based on the pitch peak position and the number of pulses. The search position of the pulse is dense near the pitch peak,
The other parts are sparsely allocated (effective when there is not enough bit allocation to search all sample points). That is, all sample points near the pitch peak position are subjected to the pulse position search, but the portions far from the pitch peak position are widened at intervals of the pulse position search such as every two samples or every three samples (for example, The search position is determined as in FIGS. 11B and 11C). Also, when the number of pulses is large, the number of bits allocated to one pulse is small, so that the interval between sparse portions is wider than when the number of pulses is small (the accuracy of the pulse position is rough). When the pitch period is short, as described in Embodiment 5, if the search range is limited to only a range slightly more than one pitch period from the first pitch peak in the subframe, the voice quality can be further improved. It is possible.

The pulse position search unit 75 determines an optimum combination of positions where pulses are to be formed based on the search position determined by the search position calculator 74. See “ITU-T STUDY GROUP15-CONTRIBUTION 152,” G.729-
CODING OF SPEECH AT 8 KBIT / S USING CONJUGATE-STRUC
TURE ALGEBRAIC-CODE-EXCITED LINEAR-PREDICTION (CS-
ACELP) ", COM 15-152-E July 1995", for example, when the number of pulses is four, the combination of i0 to i3 is determined so as to maximize Expression (2). (DN × DN) / RR DN = dn (i0) + dn (i1) + dn (i2) + dn (i3) RR = rr (i0, i0) + rr (i1, i1) + 2 × rr (i0, i1) + rr (I2, i2) + 2 × (rr (i0, i2) + rr (i1, i2)) + rr (i3, i3) + 2 × (rr (i0, i3) + rr (i1, i3) + rr (i2) i3)). . . (2) Here, dn (i) (i = 0 to 79: subframe length 8)
In the case of 0 sample), the target vector x ′ (i) of the pulse sound source component is subjected to backward filtering with the impulse response of the synthesis filter, and rr (i, i)
Is the autocorrelation matrix of the impulse response as in equation (3). The range of positions that i0, i1, i2, and i3 can take has been obtained by the search position calculator 74. Specifically, when the number of pulses is 4, FIGS. 13A to 13D
(Positions where the parts marked with arrows in the figure can be taken,
The scale value is a relative value when the pitch peak position is 0).

(Equation 1) . . . (3)

When the combination of the optimum pulse positions is determined by the pulse position search unit 75, the pulse excitation vector generated by the combination is output to the multiplier 77, multiplied by the pulse code vector gain, and
8 is output.

An adder 78 adds the adaptive code vector component and the pulse excitation vector component, and outputs the result as an excitation excitation vector.

(Embodiment 7) FIG. 14 shows a seventh embodiment of the present invention.
1 shows an embodiment of the present invention, and shows an excitation generation unit of a CELP-type speech coding apparatus having a configuration for determining the amplitude of a pulse before searching for a pulse. In FIG. 14, reference numeral 81 denotes an adaptive codebook which is composed of a buffer of a past excitation source signal and outputs an adaptive code vector to a pitch peak position calculator 82 and a multiplier 88. Reference numeral 82 denotes an external codebook by pitch analysis or adaptive codebook search. The obtained pitch period L and the adaptive code vector output from the adaptive codebook 81 are input,
A pitch peak position calculator 83 that outputs the pitch peak position to a search position calculator 84 and a pulse amplitude calculator 87. The number 83 searches for the number of pulses by using a pitch period L externally obtained by pitch analysis or adaptive codebook search. The number-of-pulses determiner 84 which outputs to the position calculator 84 is composed of a pitch period L externally obtained by pitch analysis or adaptive codebook search, the number of pulses output from the number-of-pulses determiner 83, and the pitch-peak position calculator 82. A search position calculator that receives the output pitch peak position as an input and outputs a pulse search position to a pulse position searcher 85. The search position calculator 85 calculates a pitch period L and a search position externally obtained by pitch analysis or adaptive codebook search. The pulse search position output from the calculator 84 and the pulse amplitude output from the pulse amplitude calculator 87 are input, and the The pulse position searcher 86 determines the combination of the positions where the pulses to be used for the sound sources are formed, and outputs the pulse excitation vector generated by the combination to the multiplier 89. The pulse position searcher 86 is determined by an external LPC analysis and LPC quantizer. An adder that subtracts the adaptive code vector (after gain multiplication) output from the multiplier 88 from the prediction residual signal obtained by the linear prediction filter and outputs the difference signal to the pulse amplitude calculator 87. 8
6, a pulse amplitude calculator that receives the difference signal output from 6 and outputs pulse amplitude information to a pulse position searcher 85;
An adder 9 multiplies an adaptive code vector gain by an adaptive code vector output from the adaptive code book 81 as an input.
A multiplier 89 that outputs to 0 and 86, a multiplier that multiplies the pulse excitation vector gain output from the pulse position searcher 85 as an input and outputs the result to an adder 90, and 90 is a multiplier 88 and This is an adder that adds the vectors output from 89 and outputs the result as an excitation sound source vector.

The operation of the excitation generator of the CELP-type speech coding apparatus configured as described above will be described with reference to FIG. The adaptive code vector output from adaptive codebook 81 is output to multiplier 88, multiplied by the adaptive code vector gain, and output to adders 90 and 86.

The pitch peak position calculator 82 detects a pitch peak from the adaptive code vector, and outputs the position to the search position calculator 84 and the pulse amplitude calculator 87. The detection (calculation) of the pitch peak position can be performed by maximizing the inner product of the impulse train vector arranged in the pitch period L and the adaptive code vector. In addition, by maximizing the inner product of the vector obtained by convolving the impulse response of the synthesis filter with the impulse train vector arranged in the pitch period L and the vector obtained by convolving the impulse response of the synthesis filter with the adaptive code vector, the pitch can be more accurately determined. It is also possible to detect a peak position.

The number-of-pulses determiner 83 determines the number of pulses used for the pulse sound source based on the value of the pitch period L, and outputs it to the search position calculator 84. The relationship between the number of pulses and the pitch period is predetermined by learning or statistically. For example, when the pitch period is 45 samples or less, 5 lines, when the pitch period is more than 45 samples and less than 80 samples, 4 lines, and 80 samples or more. In the case of (3), the number of each pulse is determined by the range of the pitch period value, such as three pulses. Furthermore, when the number of pulses is determined in consideration of a change in the number of pulses between consecutive subframes, a change in the pitch period L, and the like, the discontinuity between consecutive subframes and the quality of the rising portion of the voiced portion can be improved. Can be planned. Specifically, in consecutive subframes, when the number of pulses determined from the pitch period L decreases from five to three, the decrease in the number of pulses is given a hysteresis, and three pulses suddenly change from five. By avoiding a large change in the number of pulses between subframes by reducing the number of pulses to four instead of reducing the number to four, or when the pitch period L differs greatly between consecutive subframes, Since there is a high possibility that the accuracy of the pulse position is improved by reducing the number of pulses, the voice quality is improved. Therefore, when the pitch period L of the previous subframe and the pitch period L of the current subframe are significantly different, the current If the number of pulses is determined by a method such as setting the number of pulses to three irrespective of the value of the pitch period L of the subframe, voice quality can be further improved. A. When these methods are used, the influence of double pitch errors and half pitch errors in pitch analysis becomes liable to occur. Therefore, a pulse number determination method (for example, half pitch or double pitch For example, it is more effective to adopt a method of determining the continuity of the pitch period in consideration of the characteristics, and to increase the accuracy of pitch analysis as much as possible.

The search position calculator 84 determines the position where the pulse search is to be performed based on the pitch peak position and the number of pulses. The search position of the pulse is dense near the pitch peak,
The other parts are sparsely allocated (effective when there is not enough bit allocation to search all sample points). That is, all the sample points near the pitch peak position are subjected to the pulse position search, but the interval apart from the pitch peak position is widened every two samples or every three samples (for example, The search position is determined as in FIGS. 11B and 11C). Further, when the number of pulses is large, the number of bits allocated to one pulse is small, so that the interval between sparse portions is wider than when the number of pulses is small (the accuracy of the pulse position is rough). When the pitch period is short, as described in Embodiment 5, if the search range is limited to only a range slightly more than one pitch period from the first pitch peak in the subframe, voice quality can be further improved. It is.

The pulse position search unit 85 is connected to the search position determined by the search position calculator 84 and a pulse amplitude calculator 8 described later.
Based on the pulse amplitude information determined in step 7, an optimal combination of positions where pulses are to be set is determined. See “ITU-T STUDY GROUP15-CONTRIBUTION 152,” G.72
9-CODING OF SPEECH AT 8 KBIT / S USING CONJUGATE-STR
UCTURE ALGEBRAIC-CODE-EXCITED LINEAR-PREDICTION (C
S-ACELP) ", COM 15-152-E July 1995", for example, when the number of pulses is 4, the combination of i0 to i3 is determined so as to maximize the equation (4). I do. DN × DN / RR DN = a0 × dn (i0) + a1 × dn (i1) + a2 × dn (i2) + a3 × dn (i3) RR = a0 × a0 × rr (i0, i0) + a1 × a1 × rr ( i1, i1) + 2 × a0 × a1 × rr (i0, i1) + a2 × a2 × rr (i2, i2) + 2 × (a0 × a2 × rr (i0, i2) + a1 × a2 × rr (i1, i2 )) + A3 * a3 * rr (i3, i3) + 2 * (a0 * a3 * rr (i0, i3) + a1 * a3 * rr (i1, i3) + a2 * a3 * rr (i2, i3)). . . (4) Here, dn (i) (i = 0 to 79: subframe length 8)
In the case of 0 sample), the impulse response of the synthesis filter is convolved with the target vector of the pulse sound source component, and rr (i, i) is the autocorrelation matrix of the impulse response as shown in Expression (3). Also, i0, i1, i2, i3
The range of positions that can be taken by is obtained by the search position calculator 84. Specifically, when the number of pulses is four, FIG.
3 (a) to 3 (d) (possible positions indicated by arrows in the figure, and the scale values are relative values with the pitch peak position being 0). A0, a1, a2, and a3 are pulse amplitudes obtained by the pulse amplitude calculator 87.

When the combination of the optimum pulse positions is determined by the pulse position searcher 85, the pulse excitation vector generated by the combination is output to the multiplier 89, multiplied by the pulse code vector gain, and
Output to 0.

An adder 86 subtracts an adaptive code vector component (a product obtained by multiplying an adaptive code vector by an adaptive code vector gain) from a linear prediction residual signal (prediction residual vector) obtained by an externally performed LPC analysis. , And outputs the difference signal to the pulse amplitude calculator 87. In addition, CELP
In the sound source section of the adaptive speech coding apparatus, in general, the adaptive code vector gain and the noise code vector (corresponding to the pulse excitation vector in the present invention) gain are adaptive codebook search and noise codebook search (the pulse code search in the present invention). (Corresponding to position search), the vector obtained by multiplying the adaptive signal vector by the adaptive code vector gain cannot be obtained before the pulse position search. Therefore, the adder 86
The adaptive code vector component used in the subtraction is obtained by multiplying the adaptive code vector by the adaptive code vector gain (not the final optimal adaptive code vector gain) obtained from Expression (5) at the time of searching the adaptive code book.

(Equation 2) . . . (5) Here, x (n) is a so-called target vector, which is obtained by removing the zero input response of the LPC synthesis filter of the current subframe from the input signal weighted in the auditory sense. Further, y (n) is a component generated by the adaptive code vector in the synthesized speech signal. Is convoluted.

Using the pitch peak position calculated by the pitch peak position calculator 82, the pulse amplitude calculator 87 divides the difference signal output from the adder 86 into the vicinity of the pitch peak position and other parts. The average value of the power of each part or the average value of the absolute value of the signal amplitude at each sample point included in each part is obtained, and the respective amplitudes are used as the pulse amplitude near the pitch peak position and the pulse amplitude of the other parts. Output to the pulse position searcher 85. The pulse position searcher 85 evaluates Expression (4) using different amplitudes between the pulse near the pitch pulse and the pulse in the other part, and performs a pulse position search. A pulse sound source vector, which is expressed by the pulse position determined in the pulse position search and the pulse amplitude assigned to the pulse at that position, is output from the pulse position search device 85.

The adder 90 adds the adaptive code vector component and the pulse excitation vector component, and outputs the result as an excitation excitation vector.

(Embodiment 8) FIG. 15 shows an eighth embodiment of the present invention.
And shows an excitation generator of a CELP-type speech coding apparatus having a configuration in which a search position used for pulse search is switched based on a determination result of continuity of a pitch period. In FIG. 15, reference numeral 91 denotes an adaptive codebook that outputs an adaptive code vector to a pitch peak position calculator 92 and a multiplier 99. Reference numeral 92 denotes an adaptive codebook that receives the adaptive code vector output from the adaptive codebook 91 and the pitch period L as inputs. Search position calculator 94 for pitch peak position in vector
The pitch number calculator 93 outputs the pitch period L as an input, and the pulse number determiner 93 outputs the number of pulses of the pulse sound source to the search position calculator 94. The pitch number L and the pitch peak position calculator 92 A search position calculator that receives the pitch peak position output from the controller and the number of pulses output from the pulse number determiner 93 and outputs a pulse search position to a pulse position searcher 97 via a switch 98;
95 receives the pitch period L of the current subframe as an input,
A delay unit that delays by a subframe and outputs the result to the decision unit 96.
A decision unit which outputs the pitch cycle continuity determination result to the switch 98 using the pitch period of the previous subframe output from the switch 98 as an input, and a search unit 97 for searching for a pulse input from the search position calculator 94 via the switch 98 The position or fixed search position input via switch 98 and switch 9
8, a pitch position L is input, a search for a pulse position is performed using the input search position and the input pitch period L, and the pulse excitation vector is multiplied by the multiplier 100.
The pulse position searcher 98 outputs a pulse search position 98 which is switched based on the determination result input from the determiner 96. One of the switches is used to calculate the pulse search position by the search position calculator 94. Is used to switch between the searched search position and a predetermined fixed search position, and the other system of switches is used to determine whether or not to input the pitch period L to the pulse position searcher 97 or not.
Used for F. A multiplier 99 receives the adaptive code vector output from the adaptive codebook 91 as an input, multiplies the adaptive code vector gain and outputs the result to the adder 101, and 100 receives a pulse excitation vector output from the pulse position searcher 97 as an input. , Multiplying by the pulse source vector gain
Is an adder that adds the vectors input from the multipliers 99 and 100 and outputs the result as an excitation sound source vector.

The operation of the excitation generator of the CELP-type speech coding apparatus configured as described above will be described with reference to FIG. The adaptive codebook 91 is composed of a buffer of a past excitation source, extracts a corresponding portion from the buffer of the excitation source based on a pitch cycle or pitch lag obtained by an external pitch analysis or an adaptive codebook search means, and generates an adaptive codebook. The vector is output to the pitch peak position calculator 92 and the multiplier 99 as a vector. Adaptive codebook 9
The adaptive code vector output from 1 to the multiplier 99 is multiplied by the adaptive code vector gain and output to the adder 101.

The pitch peak position calculator 92 detects a pitch peak from the adaptive code vector and outputs the position to the search position calculator 94. The detection (calculation) of the pitch peak position can be performed by maximizing the inner product of the impulse train vector arranged in the pitch period L and the adaptive code vector. In addition, by maximizing the inner product of the vector obtained by convolving the impulse response of the synthesis filter with the impulse train vector arranged in the pitch period L and the vector obtained by convolving the impulse response of the synthesis filter with the adaptive code vector, the pitch can be more accurately determined. It is also possible to detect a peak position.

The number-of-pulses determiner 93 determines the number of pulses used for the pulse sound source based on the value of the pitch period L, and outputs it to the search position calculator 94. The relationship between the number of pulses and the pitch cycle is predetermined by learning or statistically. For example, when the pitch cycle is 45 samples or less, 5 is used.
Book, 4 if more than 45 samples and less than 80 samples
The number of each pulse is determined by the range of the pitch period value, such as three, for 80 samples or more.

The search position calculator 94 determines the position where the pulse search is to be performed based on the pitch peak position and the number of pulses. The search position of the pulse is dense near the pitch peak,
The other parts are sparsely allocated (effective when there is not enough bit allocation to search all sample points). That is, all the sample points near the pitch peak position are subjected to the pulse position search, but the interval apart from the pitch peak position is widened every two samples or every three samples (for example, The search position is determined as in FIGS. 11B and 11C). Further, when the number of pulses is large, the number of bits allocated to one pulse is small, so that the interval between sparse portions is wider than when the number of pulses is small (the accuracy of the pulse position is rough). When the pitch period is short, as described in Embodiment 5, if the search range is limited to only a range slightly more than one pitch period from the first pitch peak in the subframe, voice quality can be further improved. It is.

The pulse position searcher 97 determines an optimum combination of the search position determined by the search position calculator 94 or a predetermined fixed search position and a position where a pulse is to be formed based on the pitch period L. For the method of pulse search, see "ITU-T STUDY GROUP15-CONTRIBUTION 152," G.729-C
ODING OF SPEECH AT 8 KBIT / S USING CONJUGATE-STRUCT
URE ALGEBRAIC-CODE-EXCITED LINEAR-PREDICTION (CS-A
CELP) ", COM 15-152-EJuly 1995", for example, when the number of pulses is four, the combination of i0 to i3 is determined so as to maximize Expression (2).

The switching of the switch 98 is performed on the basis of the judgment result of the judgment unit 96. The determiner 96 determines whether or not the pitch periods are continuous using the pitch period L of the current subframe and the pitch period of the immediately preceding subframe input from the delay unit 95. Specifically, when the difference between the value of the pitch period of the current subframe and the value of the pitch period of the immediately preceding subframe is equal to or less than a predetermined or calculated threshold, the pitch period is considered to be continuous. judge. If it is determined that the pitch period is continuous, the current subframe is regarded as a voiced / voiced stationary part, and the switch 98 connects the search position calculator 94 and the pulse position searcher 97 to change the pitch period L to the pulse position. Searcher 9
7 (one system of the switch 98 is switched to the search position calculator 94, and the other system is turned on to input the pitch period L to the pulse position search device 97). If it is determined that the pitch cycle is not continuous (the difference between the pitch cycle of the current subframe and the pitch cycle of the immediately preceding subframe exceeds a threshold), the current subframe is voiced.
Not voiced stationary part (voiceless part / voiced rising part)
The switch 98 inputs a predetermined fixed search position to the pulse searcher 97 and does not input the pitch period L to the pulse position searcher (one system of the switch 98 is switched to the fixed search position, and One system is OF
In the F state, the pitch period L is not input to the pulse position search unit 97).

When the combination of the optimum pulse positions is determined by the pulse position search unit 97, the pulse excitation vector generated by the combination is output to the multiplier 100, multiplied by the pulse code vector gain, and added to the adder 101. Is output.

The adder 101 adds the adaptive code vector component and the pulse excitation vector component, and outputs the result as an excitation excitation vector.

The table shown in FIG. 16 shows an example of the contents of the fixed search position in FIG. FIG. 16B shows a case where the search positions are fixed so that the search positions are evenly scattered over the entire subframe when eight positions are assigned per pulse in the same manner as the search positions shown in FIG. (Rather than densely near the pitch peak and sparsely in other parts, the density is made uniform as a whole). FIG. 16 (a) shows that, instead of reducing the search positions assigned to two of the four pulses to four, four search positions are used and all the sample points in the sub-frame are located at one of the four positions. It is included in the search position group (the number of bits for expressing the pulse position is (a)
13 (b) and FIG. 13). In this case, as shown in FIG. 16B, there are no positions that are not searched at all, so that the performance of FIG. 16A generally improves with the same number of bits.

In this embodiment, the pulse number determiner 93 is used.
Although the excitation generation unit of the variable pulse number speech coding apparatus having the above is shown, the pulse search position switching using the continuity of the pitch period can be performed even in the fixed pulse number type without the pulse number determiner 93. It is valid. Further, in the present embodiment, the continuity of the pitch cycle is determined only from the pitch cycle of the immediately preceding subframe and the current subframe. However, the determination accuracy is further improved by using the pitch cycle of the past subframe. It is also possible to make it.

(Embodiment 9) FIG. 17 shows a ninth embodiment of the present invention.
And the pitch gain (adaptive code vector gain) has a two-stage quantization configuration.
The first-stage target is a pitch gain calculated immediately after the adaptive codebook search, and has a configuration in which a search position used for pulse search is switched based on the first-stage quantization pitch gain. Is shown. In FIG. 17, reference numeral 111 denotes an adaptive codebook that outputs an adaptive code vector to a pitch peak position calculator 112, a pitch gain calculator 116, and a multiplier 123.
Is a pitch peak position calculator that receives the adaptive code vector and the pitch period L output from the adaptive codebook 111 and outputs a pitch peak position in the adaptive code vector to the search position calculator 114. As an input, a pulse number determiner that outputs the number of pulses of the pulse sound source to the search position calculator 114. The pitch period L and the pitch peak position output from the pitch peak position calculator 112 are output from the pulse number determiner 113. A search position calculator that outputs the pulse search position to the pulse position searcher 119 via the switch 115 with the number of pulses input as input,
Reference numeral 15 denotes an interlocking two-system switch that switches based on the judgment result input from the judgment unit 118. One of the switches is configured to set the pulse search position to the search position calculated by the search position calculator 114 in advance. The switch of the other system is used for ON / OFF of whether or not to input the pitch period L to the pulse position search unit 119. A pitch gain calculator 116 receives the adaptive code vector output from the adaptive codebook 111, the target vector of the current frame, and an impulse response as input, and outputs a pitch gain to a quantizer 117. 117 outputs from the pitch gain calculator 116. Quantizer 118 which quantizes the pitch gain to be output and outputs the quantized pitch gain to determiner 118 and adders 120 and 122.
The switch 1 receives the first-stage quantization pitch gain output from the quantizer 117 and determines the pitch periodicity determination result using the switch 1.
A decision unit 119 for outputting to the reference numeral 15 is a pulse search position input from the search position calculator 114 via the switch 115 or a fixed search position input via the switch 115, and a pitch input via the switch 115. Period L
, A pulse position searcher that searches for a pulse position using the input search position and the pitch period L, and outputs a pulse excitation vector to the multiplier 124.
Is input with the first-stage quantization pitch gain output from the quantizer 117 and the difference quantization pitch gain output from the difference quantizer 121, and multiplies the addition result as an optimum quantization pitch gain (adaptive code vector gain). An adder 121 outputs the difference value output from the adder 122, and a difference quantizer 121 outputs the quantized value to the adder 120. The adder 122 determines the adaptive code vector and the pulse excitation vector. Pitch gain (adaptive code vector gain) calculated externally after quantization and a quantizer 117
The adder 123 receives the first-stage quantization pitch gain (adaptive code vector gain) output from the input device and outputs a difference between them to a difference quantizer 121. The adder 123 inputs the adaptive code vector output from the adaptive codebook 111. And the adder 12
A multiplier that multiplies the quantization pitch gain (adaptive code vector gain) output from 0 and outputs the result to the adder 125;
4 is a multiplier which receives the pulse excitation vector output from the pulse position searcher 119 as input, multiplies the pulse excitation vector gain and outputs the result to the adder 125, and 125 is a multiplier 123
And an adder that adds the vectors input from the and 124 and outputs the result as an excitation sound source vector.

The operation of the sound source generation unit of the speech coding apparatus configured as described above will be described with reference to FIG. The adaptive codebook 111 is composed of a buffer of a past excitation source, extracts a corresponding portion from the buffer of the excitation source based on the pitch period or pitch lag obtained by an external pitch analysis or an adaptive codebook search means, Pitch peak position calculator 1 as vector
12 and pitch gain calculator 116 and multiplier 12
Output to 3. The adaptive code vector output from the adaptive codebook 111 to the multiplier 123 is multiplied by the quantization pitch gain (adaptive code vector gain) output from the adder 120 and output to the adder 125.

The pitch peak position calculator 112 detects the pitch peak from the adaptive code vector and outputs the position to the search position calculator 114. The detection (calculation) of the pitch peak position can be performed by maximizing the inner product of the impulse train vector arranged in the pitch period L and the adaptive code vector. In addition, by maximizing the inner product of the vector obtained by convolving the impulse response of the synthesis filter with the impulse train vector arranged in the pitch period L and the vector obtained by convolving the impulse response of the synthesis filter with the adaptive code vector, the pitch can be more accurately determined. It is also possible to detect a peak position.

The number-of-pulses determiner 113 determines the number of pulses used for the pulse sound source based on the value of the pitch period L, and outputs it to the search position calculator 114. The relationship between the number of pulses and the pitch period is predetermined by learning or statistically. For example, when the pitch period is 45 samples or less, 5 lines, when the pitch period is more than 45 samples and less than 80 samples, 4 lines, and 80 samples or more. In the case of (3), the number of each pulse is determined by the range of the pitch period value, such as three pulses.

The search position calculator 114 determines the position where the pulse search is to be performed based on the pitch peak position and the number of pulses. Pulse search positions are distributed so as to be dense near the pitch peak and sparse in other portions (effective when there is not enough bit allocation to search all sample points). That is, all sample points near the pitch peak position are subject to pulse position search,
In a portion away from the pitch peak position, the interval of the pulse position search is widened such as every two samples or every three samples (for example, the search position is determined as shown in FIGS. 11B and 11C). Further, when the number of pulses is large, the number of bits allocated to one pulse is small, so that the interval between sparse portions is wider than when the number of pulses is small (the accuracy of the pulse position is rough). When the pitch period is short, as described in Embodiment 5, if the search range is limited to only a range slightly more than one pitch period from the first pitch peak in the subframe, the voice quality can be further improved. It is possible.

The pulse position search unit 119 determines an optimum combination of the search position determined by the search position calculator 114 or a predetermined fixed search position and a position where a pulse is to be formed based on the pitch period L. See “ITU-T STUDY GROUP15-CONTRIBUTION 152,” G.72
9-CODING OF SPEECH AT 8 KBIT / S USING CONJUGATE-STR
UCTURE ALGEBRAIC-CODE-EXCITED LINEAR-PREDICTION (C
S-ACELP) ", COM 15-152-E July 1995", for example, when the number of pulses is four, the combination of i0 to i3 is determined so as to maximize Expression (2). I do.

The switching of the switch 115 is performed by the decision unit 11.
The determination is performed based on the determination result of Step 8. The determiner 118 uses the first-stage quantization pitch gain output from the quantizer 117 to determine whether the current subframe is a subframe with a strong pitch periodicity. Specifically, when the first-stage quantization pitch gain is within a predetermined range or a range obtained by calculation, it is determined that the pitch periodicity is strong.
If it is determined that the pitch periodicity is strong, the current subframe is regarded as a voiced / voiced stationary part, and the switch 115 connects the search position calculator 114 and the pulse position searcher 119 to determine the pitch period L as the pulse position search. (One system of the switch 115 is switched to the search position calculator 114, and the other system is turned on to input the pitch period L to the pulse position search device 119). If it is determined that the pitch cycle is not continuous (the difference between the pitch cycle of the current subframe and the pitch cycle of the immediately preceding subframe exceeds a threshold), the current subframe is not a voiced / voiced stationary part (unvoiced part / voiced rising) The switch 115 inputs a predetermined fixed search position to the pulse searcher 119, and does not input the pitch period L to the pulse position searcher (one system of the switch 115 is set to the fixed search position). The other system is switched off, and the pitch period L is not input to the pulse position search unit 119).

When the combination of the optimum pulse positions is determined by the pulse position searcher 119, the pulse excitation vector generated by the combination is multiplied by the multiplier 124.
Are multiplied by the pulse code vector gain and output to the adder 125.

The pitch gain calculator 116 uses the impulse response of the filter obtained by cascading the quantized LPC synthesis filter and the perceptual weighting filter of the current subframe, the target vector, and the adaptive code vector output from the adaptive codebook. , Equation (5) is used to calculate a pitch gain (adaptive code vector gain). The calculated pitch gain is quantized by the quantizer 117 and output to the determiner 118 for determining the strength of the pitch periodicity and to the adders 120 and 122. The adder 122 outputs the optimum quantization pitch gain calculated after all of the excitation codebook search (the adaptive codebook search and the noise codebook search (the pulse position search in this embodiment)) and the output from the quantizer 117. (First stage) and a quantized pitch gain, and calculates a difference quantizer 121.
Output to The difference value quantized by the difference quantizer 121 is added by the adder 120 to the first-stage quantization pitch gain output from the quantizer 117, and is output to the multiplier 123 as the optimum quantization pitch gain. .

Multiplier 123 multiplies the adaptive code vector output from adaptive codebook 111 by the optimum quantization pitch gain, and outputs the result to adder 125.

The adder 125 adds the adaptive code vector component and the pulse excitation vector component, and outputs the result as an excitation excitation vector.

In this embodiment, the first-stage quantization pitch gain of the current subframe is used as an input to the decision unit 118. However, when general gain quantization is used (as shown in this embodiment). In the case where multi-stage quantization is not used), the quantization pitch gain (adaptive code vector gain) of the immediately preceding subframe can be used as an input to the determiner 118. Also, in the present embodiment, the excitation generator of the variable pulse number speech coding apparatus having the pulse number determiner has been described.
Even in a fixed pulse number type having no pulse number determiner, it is effective to switch the pulse search position by determining the strength of the periodicity using the pitch gain value.

(Embodiment 10) FIG. 18 shows the first embodiment of the present invention.
0, and uses a continuity of the phase of the excitation signal waveform between consecutive subframes to switch the phase adaptation process for the noise codebook in the backward direction. Show. In FIG. 18, 180
Reference numeral 1 denotes an adaptive code vector for calculating a pitch peak position calculator 180.
2 and an adaptive codebook 1802 to be output to the multiplier 1810, and the adaptive codebook 1802 receives the adaptive code vector output from the adaptive codebook 1801 and the pitch period L as inputs and determines the pitch peak position in the adaptive code vector by the delay unit 1803 and the decision unit 1806. A pitch peak position calculator 1803 that outputs the pitch peak position output from the pitch peak position calculator 1802 to the search position calculator 1807, and delays it by one subframe and outputs it to the pitch peak position predictor 1805 A delay device 1804 receives a pitch period L as an input,
Pitch peak position estimator 1 delayed by one subframe
The delay unit 1805 outputs the pitch peak position in the immediately preceding subframe output from the delay unit 1803, the pitch period in the immediately preceding subframe output from the delay unit 1804, and the pitch period L in the current subframe. A pitch peak position predictor that outputs a predicted pitch peak position to a determiner 1806 as an input;
06 receives the pitch peak position output from the pitch peak position calculator 1802 and the predicted pitch peak position output from the pitch peak position estimator 1805, and determines the phase continuity between the immediately preceding subframe and the current subframe. It is determined whether or not there is a switch.
, And 1807 receives the pitch peak position and the pitch period L output from the pitch peak position calculator 1802 as inputs, and switches the search position of the sound source pulse to the switch 1.
A search position calculator that outputs to the pulse position searcher 1809 via 808, a switch 1808 that switches based on the determination result output from the determiner 1806, and is set in advance as the search position output from the search position calculator. It is used for switching between fixed search positions. Reference numeral 1809 designates a search position of a sound source pulse input from the search position calculator 1807 via the switch 1808 or a fixed search position input via the switch 1808, and a pitch period L, respectively. A pulse position searcher that searches for the position of the excitation pulse by using the pulse period L and the pulse excitation vector, and outputs the pulse excitation vector to the multiplier 1812. The adaptive position detection unit 1810 receives the adaptive code vector output from the adaptive codebook 1801 as an input, and A multiplier 1812 which multiplies by a vector gain and outputs it to an adder 1811; a multiplier 1812 which receives the pulse excitation vector output from the pulse position searcher 1809 as an input, multiplies by a quantized pulse excitation vector gain, and outputs it to an adder 1811; Are the vectors output from the multipliers 1810 and 1812, respectively. As input, performs addition of the input vector, an adder which outputs as excitation excitation vector.

The operation of the sound source generation unit of the speech coding apparatus configured as described above will be described with reference to FIG. The adaptive codebook 1801 is constituted by a buffer of a past excitation source, extracts a corresponding part from the buffer of the excitation source based on a pitch cycle or pitch lag obtained by an external pitch analysis or an adaptive codebook search means, The vector is output to the pitch peak position calculator 1802 and the multiplier 1810 as a vector. The adaptive code vector output from the adaptive codebook 1801 to the multiplier 1810 is multiplied by a quantized adaptive code vector gain quantized by an external gain quantizer to obtain an adder 1811.
Is output to

A pitch peak position calculator 1802 detects a pitch peak from the adaptive code vector, and determines the position of the pitch peak by using a delay unit 1803, a determiner 1806, and a search position calculator 1802.
7 is output. The detection (calculation) of the pitch peak position can be performed by maximizing the normalized cross-correlation function between the impulse train vector arranged in the pitch period L and the adaptive code vector. Also, maximizing a normalized cross-correlation function between a vector obtained by convolving the impulse response of the synthesis filter with the impulse train vector arranged in the pitch period L and a vector obtained by convolving the impulse response of the synthesis filter with the adaptive code vector. Accordingly, the pitch peak position can be detected with higher accuracy.
Furthermore, if post-processing is performed to set the position where the amplitude value becomes maximum among the one-pitch periodic waveforms including the detected pitch-peak position as the pitch peak, erroneous detection of the second peak in the one-pitch periodic waveform is avoided. It is also possible.

The delay unit 1803 delays the pitch peak position calculated by the pitch peak position calculator 1802 by one subframe, and delays the pitch peak position by a subframe.
Output to That is, the pitch peak position in the immediately preceding subframe is input from the delay unit 1803 to the pitch peak position predictor 1805. Delay device 1804 delays pitch period L by one subframe and outputs the result to pitch peak position calculator 1805. That is, the pitch period in the immediately preceding subframe is input from the delay unit 1804 to the pitch peak position predictor 1805.

The pitch peak position predictor 1805 calculates the pitch peak position in the subframe immediately before input from the delay unit 1803, the pitch period in the subframe immediately before input from the delay unit 1804, and the pitch in the current subframe. With the cycle L as an input, a pitch peak position in the current subframe is predicted, and the predicted pitch peak position is output to the determiner 1806. The predicted pitch peak position is obtained by Expression (6) (see FIG. 19). Φ (N) = Φ (N−1) + n × T (N−1) + T (N) −L, n = INT ((L−Φ (N−1)) / T (N−1)). . . (6)

In the above equation, Φ (k) represents the first pitch peak position in the k-th subframe with the beginning of the subframe being 0, and T (k) is the k-th subframe. Is a pitch cycle of a sound source (sound) signal in L, and L is a subframe length. Also, n
Is an integer indicating how many pitch period lengths are included from the first pitch peak position (Φ (k)) in the k-th subframe to the end of the k-th subframe (rounded down to the decimal point). It is a numerical value. (K = 0,1,
2, ...)

The determiner 1806 receives the pitch peak position output from the pitch peak position calculator 1802 and the predicted pitch peak position output from the pitch peak position predictor 1805 as inputs, and determines the pitch peak position as the predicted pitch peak position. If not significantly different, it is determined that the phases are continuous. If the pitch peak position is significantly different from the predicted pitch peak position, it is determined that the phases are not continuous. Then, the determination result is output to the switch 1808. When comparing the pitch peak position with the predicted pitch peak position, when the pitch peak position or the predicted pitch peak position exists near the subframe boundary, the possibility that the position one pitch cycle later is the pitch peak position is also considered. Then, the continuity of the phase is determined by comparing the pitch peak position with the predicted pitch peak position.

The search position calculator 1807 determines the search position of the sound source pulse based on the pitch peak position, and sets the search position via the switch 1808 to the pulse position searcher 18.
09 is output. As a method of determining the search position, for example, as described in the sixth and eighth embodiments, the search position is determined so that the vicinity of the pitch peak is densely distributed and the other portions are sparsely distributed. It is also effective to change the number of sound source pulses using the pitch period information or limit the search range of the sound source pulses as described in the sixth and eighth embodiments.

The switch 1808 switches between performing a phase-adaptive excitation pulse search or a fixed-position excitation pulse search (or a general noise codebook search) based on the determination result of the determiner 1806. . That is,
When the determination result of the determiner 1806 is “with phase continuity”, the search position calculator 1807 and the pulse position searcher 18
09 and connects the sound source pulse search position calculated by the search position calculator 1807 to the pulse position searcher 1809.
(That is, a phase-adaptive sound source pulse search is performed). Conversely, the determination result of the determiner 1806 is
In the case of "no phase continuity", switching is performed so that the fixed search position is input to the pulse position searcher 1809 (when switching from the general noise codebook search, a configuration is provided with a separate noise codebook searcher. To be used by switching to the device 1809).

The pulse position searcher 1809 uses the sound source pulse search position determined by the search position calculator 1807 or a predetermined fixed search position, and a position at which a sound source pulse is to be formed using the pitch period L input separately. To determine the optimal combination of See ITU-T Reco
mmendation G.729: Coding of Speech at 8 kbits / susi
ng Conjugate-Structure Algebraic-Code-Excited Line
ar-Prediction (CS-ACELP), March 1996 ”, for example, when the number of pulses is four, the values of i0 to i3 are set so as to maximize the expression (2) shown in the sixth embodiment. Determine the combination. Note that the polarity of each excitation pulse at this time is the target vector of the noise codebook component, that is, a signal obtained by subtracting the zero input response signal of the auditory weighting synthesis filter and the signal of the adaptive codebook component from the auditory weighted input speech. Is determined before performing the pulse position search so as to be equal in polarity at each position of the vector. When the pitch period is shorter than the subframe length, a pitch period filter is applied as described in the fifth embodiment, so that the excitation pulse is not an impulse but a pulse train of the pitch period. In the case of performing such a pitch periodization process, if the pitch periodization filter is applied to the impulse response vector of the auditory weighting synthesis filter in advance, the maximization of Expression (2) is performed in the same manner as the case where the pitch periodization is not performed. Thus, a search for a sound source pulse can be performed. A pulse is generated at the position of each sound source pulse determined in this way according to the polarity of each determined sound source pulse, and if a pitch period filter is applied using the pitch period L, a pulse sound source vector is generated. The generated pulse excitation vector is output to multiplier 1812.
The pulse excitation vector output from the pulse position searcher 1809 to the multiplier 1812 is multiplied by a quantized pulse excitation vector gain quantized by an external gain quantizer and output to the adder 1811.

The adder 1811 performs vector addition of the adaptive code vector component output from the multiplier 1810 and the pulse excitation vector component output from the multiplier 1812, and outputs the result as an excitation excitation vector.

In the speech coding apparatus according to the present invention, a state where the fixed search position is continuously selected is apt to occur in portions other than the voiced stationary portion, so that the reset is performed when the influence of the transmission path error propagates. Can be obtained. (In the case where the pulse position is expressed by a relative position where the pitch peak position is 0, if a transmission path error occurs once and the contents of the adaptive codebooks on the encoder side and the decoder side greatly differ, the following frame is used. Even if there is no transmission path error, a phenomenon may occur in which the pitch peak position continues to be inconsistent between the encoder side and the decoder side, and the effect of the error is dragged for a long time.

As for the method of setting the pulses, when a fixed number of pulses, for example, four pulses are set in a search range, for example, somewhere in 32 positions, the 32 positions are divided into four as described above, and one pulse is set. All of the combinations (8 × 8
In addition to the method of searching (× 8 × 8 ways), there is a method of searching for all the combinations that select four places from 32 places. In addition to the combination of the impulses having the amplitude of 1, a combination of a plurality of pulses, for example, two pulses, or a combination of impulses having different amplitudes may be used.

(Embodiment 11) FIG. 20 shows a first embodiment of the present invention.
1 shows an embodiment, and shows a sound source generation unit of a CELP-type speech coding apparatus that switches between performing and not performing phase adaptation processing depending on whether or not a strong pulse property exists in the shape of an adaptive code vector. ing. In FIG. 20, reference numeral 2001 denotes an adaptive codebook that outputs an adaptive code vector to a pitch peak position calculator 2002, a pulse determination unit 2003, and a multiplier 2007. 2002 denotes an adaptive code vector output from the adaptive codebook 2001 and a pitch period L. And a pitch peak position calculator that outputs the pitch peak position in the adaptive code vector to the pulse property determiner 2003 and the search position calculator 2004, and the adaptive code vector and the adaptive code vector 2003 output from the adaptive codebook 2001. Using the pitch peak position output from the peak position calculator 2002 and the pitch period L input from the outside as inputs, it is determined whether or not the adaptive code vector has a good pulse property, and the determination result is output to the switch 2005. The sex determiner 2004 has a pitch period L and a pitch peak that are input from the outside. As input pitch peak position output from the position calculator 2002, search position calculator which outputs the search position of the sound source pulse in the pulse position searcher 2006 through the switch 2005,
A switch 2005 is switched based on the determination result output from the pulse determination unit 2003, and is used to switch between the search position output from the search position calculator 2004 and a predetermined fixed search position. Reference numeral 2006 denotes a search position of a sound source pulse input from the search position calculator 2004 via the switch 2005 or a switch 2005.
A fixed search position input through the input unit and a pitch period L input from the outside are input, and the input sound source pulse search position and the pitch period L are used to search for the position of the sound source pulse, and the pulse sound source vector is multiplied. A pulse position searcher that outputs to the adder 2009, a multiplier that receives the adaptive code vector output from the adaptive codebook 2001 as an input, multiplies the quantized adaptive code vector gain, and outputs the result to the adder 2008, and 2009 a pulse position searcher A multiplier that receives the pulse excitation vector output from 2006 as an input, multiplies the quantized pulse excitation vector gain, and outputs the result to the adder 2008. The input 2008 receives the vectors output from the multipliers 2007 and 2009, respectively, and receives the input vector. Is an adder that performs the addition and outputs as an excitation sound source vector.

The operation of the sound source generating section of the speech coding apparatus configured as described above will be described with reference to FIG. The adaptive codebook 2001 includes a buffer of a past excitation source, extracts a corresponding portion from a buffer of the excitation source based on a pitch period or pitch lag obtained by an external pitch analysis or an adaptive codebook search means, and extracts an adaptive code. As a vector, a pitch peak position calculator 2002, a pulse property determiner 2003, and a multiplier 2
007. Multiplier 20 from adaptive codebook 2001
The adaptive code vector output to 07 is multiplied by a quantized adaptive code vector gain quantized by an external gain quantizer and output to an adder 2008.

The pitch peak position calculator 2002 detects a pitch peak from the adaptive code vector, and outputs the position to each of the pulse determination unit 2003 and the search position calculator 2004. The detection (calculation) of the pitch peak position
This can be performed by maximizing the normalized cross-correlation function between the impulse train vector and the adaptive code vector arranged in the pitch period L. Also, maximizing a normalized cross-correlation function between a vector obtained by convolving the impulse response of the synthesis filter with the impulse train vector arranged in the pitch period L and a vector obtained by convolving the impulse response of the synthesis filter with the adaptive code vector. Accordingly, the pitch peak position can be detected with higher accuracy. further,
If post-processing is performed to set the position where the amplitude value becomes maximum from the one-pitch periodic waveform including the detected pitch-peak position as the pitch peak, it is possible to avoid erroneous detection of the second peak in the one-pitch periodic waveform. It is possible.

The pulse property determiner 2003 determines whether or not the signal power of the adaptive code vector is concentrated near the pitch peak position calculated by the pitch peak position calculator 2002. The determination result of "with pulse property" is output to the switch 2005, and the determination result of "no pulse property" is output to the switch 2005 when no signal power concentration is observed. As a method for checking whether or not the signal power is concentrated, for example, the following method can be considered. First, an adaptive code vector having one pitch cycle length including a pitch peak position is cut out, the power of the whole cut out signal is calculated, and this is set to PW0. Next, an adaptive code vector having a length of one-half to one-third of the pitch in the vicinity of the pitch peak position is cut out, and the cut-out signal power is calculated to be PW1. PW1 / PW
When the value of 0 is equal to or more than a predetermined value (for example, about 0.5 to 0.6), since the signal power is concentrated near the pitch peak, it is possible to determine that the pulse property is high.
As another determination method, an error between the impulse sequence vector and the adaptive code vector when the adaptive code vector is approximated by an impulse sequence vector arranged at a pitch cycle interval in which the first impulse is at a pitch peak position is used. There is a judgment method that was used. Further, maximizing a normalized cross-correlation function between a vector obtained by convolving the impulse response of the synthesis filter with the impulse train vector arranged at the pitch period L and a vector obtained by convolving the impulse response of the synthesis filter with the adaptive code vector. When the pitch peak position is obtained by the above, an error between a vector obtained by convolving the impulse response of the synthesis filter with the impulse train vector arranged in the pitch period L and a vector obtained by convolving the impulse response of the synthesis filter with the adaptive code vector is obtained. There is a determination method using As means for evaluating the error between these vectors, a prediction gain as shown in equation (7), a normalized cross-correlation function as shown in equation (8), and the like are used. Equation (7)
In (8) and (8), x (n) is the adaptive code vector or a vector obtained by convolving the adaptive code vector with the impulse response of the synthesis filter, and y (n) is the impulse train vector or the impulse response of the synthesis filter convolved with the impulse train vector. This is the embedded vector. In either case, if the value is, for example, 0.3 to 0.4 or more, it can be determined that the adaptive code vector has a somewhat strong pulse property.

(Equation 3) . . . (7)

(Equation 4) . . . (8)

The search position calculator 2004 determines the search position of the sound source pulse with reference to the pitch peak position, and sets the search position to the pulse position searcher 20 via the switch 2005.
06 is output. As a method of determining the search position, for example, as described in the sixth and eighth embodiments, the search position is determined so that the vicinity of the pitch peak is densely distributed and the other portions are sparsely distributed. It is also effective to change the number of sound source pulses using the pitch period information or limit the search range of the sound source pulses as described in the sixth and eighth embodiments.

The switch 2005 is connected to the pulse
It is to switch between performing a phase-adaptive sound source pulse search or performing a sound source pulse search at a fixed position based on the determination result of 03. That is, when the determination result of the pulse property determiner 2003 is “with pulse property”, the search position calculator 2004 and the pulse position searcher 2006 are connected, and the sound source pulse search position calculated by the search position calculator 2004 is determined. The signal is input to the pulse position search unit 2006 (that is, a phase-adaptive sound source pulse search is performed). Conversely, when the determination result of the pulse property determination unit 2003 is “no pulse property”, the fixed search position is set to the pulse position search unit 200.
6 is switched to input.

The pulse position searcher 2006 uses the sound source pulse search position determined by the search position calculator 2004 or a predetermined fixed search position, and a position at which a sound source pulse is formed using a pitch period L separately input. To determine the optimal combination of See ITU-T Reco
mmendation G.729: Coding of Speech at 8 kbits / susi
ng Conjugate-Structure Algebraic-Code-Excited Line
ar-Prediction (CS-ACELP), March 1996 ”, for example, when the number of pulses is four, the values of i0 to i3 are set so as to maximize the expression (2) shown in the sixth embodiment. Determine the combination. Note that the polarity of each excitation pulse at this time is the target vector of the noise codebook component, that is, a signal obtained by subtracting the zero input response signal of the auditory weighting synthesis filter and the signal of the adaptive codebook component from the auditory weighted input speech. Is determined before performing the pulse position search so as to be equal in polarity at each position of the vector. When the pitch period is shorter than the subframe length, a pitch period filter is applied as described in the fifth embodiment, so that the excitation pulse is not an impulse but a pulse train of the pitch period. In the case of performing such pitch periodization processing, if a pitch periodization filter is previously applied to the impulse response vector of the auditory weighting synthesis filter, the maximum value of Expression (2) can be obtained in the same manner as when pitch periodization is not performed. The search for the sound source pulse can be performed by the conversion. A pulse is generated at the position of each sound source pulse determined in this way according to the polarity of each determined sound source pulse, and if a pitch period filter is applied using the pitch period L, a pulse sound source vector is generated. The generated pulse excitation vector is output to multiplier 2009. The pulse excitation vector output from the pulse position searcher 2006 to the multiplier 2009 is multiplied by a quantized pulse excitation vector gain quantized by an external gain quantizer, and output to the adder 2008.

The adder 2008 performs vector addition of the adaptive code vector component output from the multiplier 2007 and the pulse excitation vector component output from the multiplier 2009, and outputs the result as an excitation excitation vector.

In the speech coding apparatus according to the present invention, a state where the fixed search position is continuously selected is apt to occur in portions other than the voiced stationary portion, so that the reset is performed when the influence of the transmission path error is propagated. Can be obtained. (In the case where the pulse position is expressed by a relative position where the pitch peak position is 0, if a transmission path error occurs once and the contents of the adaptive codebooks on the encoder side and the decoder side greatly differ, the following frame is used. Even if there is no transmission path error, a phenomenon may occur in which the pitch peak position continues to be inconsistent between the encoder side and the decoder side, and the effect of the error is dragged for a long time.

As for the method of setting pulses, when a fixed number of pulses, for example, four pulses are set in a search range, for example, somewhere in 32 positions, the 32 positions are divided into four as described above and one pulse is set. All of the combinations (8 × 8
In addition to the method of searching (× 8 × 8 ways), there is a method of searching for all the combinations that select four places from 32 places. In addition to the combination of the impulses having the amplitude of 1, a combination of a plurality of pulses, for example, two pulses, or a combination of impulses having different amplitudes may be used.

(Embodiment 12) FIG. 21 shows a first embodiment of the present invention.
2 shows the second embodiment, comprising an index updating means for changing the index of the pulse search position, and the encoder side of the CELP type speech coding apparatus for determining the search range of the pulse position by the pitch period and the pitch peak position of the adaptive code vector. Is shown. More specifically, in a CELP-type speech coding apparatus that performs excitation pulse search at a relative position from a pitch peak position, a pulse position is indexed sequentially from the head of a subframe, so that in a certain frame, FIG. 6 shows a sound source generation unit configured to prevent the influence of a generated transmission path error from propagating to a subsequent frame having no transmission path error. FIG.

In FIG. 21, reference numeral 2101 denotes an adaptive codebook for storing a past excitation source vector and outputting the selected adaptive code vector to a pitch peak position calculator 2102 and a pitch gain multiplier 2106. Reference numeral 2102 denotes an adaptive codebook 2101. Calculates the pitch peak position by using the adaptive code vector and pitch period L output from the as input, and outputs the pitch peak position calculator 2103 to the search position calculator 2103.
103 calculates a range in which to search for a pulse sound source using the pitch peak position and the pitch period L output from the pitch peak position calculator 2102 as inputs, and
A search position calculator 2104 for outputting to the 104 is a pulse position searcher 2105 that changes the index of each position of each sound source pulse output from the search position calculator 2103.
The index updating means 2105 outputs the search position (the index indicating the pulse position is re-assigned) output from the index updating means 2104 and the pitch period L separately calculated outside the sound source generation unit. The pulse position searcher 2106 outputs a pulse excitation vector to the pulse excitation gain multiplier 2107 and outputs an index representing the pulse excitation vector to the outside of the excitation generation unit as an encoded output. A multiplier 2107 multiplies the adaptive code vector output from 2101 by an adaptive code vector gain and outputs the result to an adder 2108. A multiplier 2107 multiplies the pulse excitation vector output from the pulse position searcher 2105 by the pulse excitation vector gain, and 2108 is the output from the multiplier 2106 Receives the output from the multiplier 2107, an adder which outputs as excitation excitation vector by vector addition.

The operation of the sound source generator configured as described above will be described with reference to FIGS. 21 and 22. In FIG. 21, adaptive codebook 2101 cuts out an adaptive code vector by a subframe length from a point that has been advanced in the past by a pitch period L calculated in advance outside the excitation generation unit, and outputs it as an adaptive code vector. If the pitch period L is less than the subframe length, a vector obtained by repeatedly connecting the extracted pitch period L until the subframe length is reached is output as an adaptive code vector.

Pitch peak position calculator 2102 uses the adaptive code vector output from adaptive code book 2101 to determine the position of the pitch peak existing in the adaptive code vector. The position of the pitch peak can be determined by maximizing the normalized cross-correlation between the impulse train arranged in the pitch cycle and the adaptive code vector. Further, it is possible to obtain the impulse trains arranged in the pitch cycle through a synthesis filter, and to minimize the error caused by passing the adaptive code vector through the synthesis filter, thereby obtaining the error more accurately.

The search position calculator 2103 determines the search position of the excitation pulse with reference to the pitch peak position, and outputs it to the index updating means 2104. As a method of determining the search position, for example, as shown in the fifth and sixth embodiments, the search position is determined so that the vicinity of the pitch peak is densely distributed and the other portions are sparsely distributed. It is also effective to change the number of sound source pulses using the pitch period information or limit the search range of the sound source pulses as described in the sixth and eighth embodiments.
Examples of specific search positions determined by the search position calculator 2103 are shown in FIGS. 10, 11B, 11C, and 13. FIG. For example, FIG. 10 specifically shows a method of limiting the pulse position search range densely in the vicinity of the pitch pulse position and sparsely in other portions. This limiting method is based on a statistical result in which positions where the pulse is likely to be raised are concentrated near the pitch pulse. If the pulse position search range is not limited, the probability that a pulse will be raised near the pitch pulse in a voiced part will be higher than the probability that it will be raised in other parts. The search position calculator calculates the sound source pulse search position using the relative position from the pitch peak position. At this time, the numerical value of the relative position where the pitch peak position is 0 is small. (See FIG. 22. FIG. 22 shows a case corresponding to FIG. 13 (a) where the number of pulses is four.)

The index updating means 2104 sets the sound source pulse search position indexed in ascending order of relative position from the pitch peak position (relative position in FIG. 22) to the absolute position where the head of the subframe is 0. After conversion into the position, the index is re-indexed in ascending order of the absolute position (absolute position in FIG. 22) and output to the pulse position searcher 2105. In this way, when the pitch peak positions calculated on the encoder side and the decoder side are different due to a transmission path error or the like, the pulse position shift can be reduced.

The pulse position search unit 2105 generates a sound source pulse using the pitch period L input separately from the sound source pulse search position where the index update unit 2104 re-indexes the search position. To determine the optimal combination of See `` ITU-T Recommendation G.729: Coding of Speech at
8 kbits / s using Conjugate-Structure Algebraic-Code
-Excited Linear-Prediction (CS-ACELP), March 1996
", For example, when the number of pulses is four, the combination of i0 to i3 is determined so as to maximize the expression (2) shown in the sixth embodiment. Note that the polarity of each excitation pulse at this time is the target vector of the noise codebook component, that is, a signal obtained by subtracting the zero input response signal of the auditory weighting synthesis filter and the signal of the adaptive codebook component from the auditory weighted input speech. If the pulse position is determined in advance before performing the pulse position search so as to be equal to the polarity at each position of the vector, the calculation amount for the search can be greatly reduced. When the pitch period is shorter than the subframe length, the fifth embodiment
By applying a pitch period filter as described above, the sound source pulse is not an impulse but a pulse train having a pitch period. In the case of performing such pitch periodization processing, if a pitch periodization filter is previously applied to the impulse response vector of the auditory weighting synthesis filter, the maximum value of Expression (2) can be obtained in the same manner as when pitch periodization is not performed. The search for the sound source pulse can be performed by the conversion. A pulse is generated at the position of each sound source pulse determined in this way according to the polarity of each determined sound source pulse, and if a pitch period filter is applied using the pitch period L, a pulse sound source vector is generated. The generated pulse excitation vector is output to multiplier 2107.
The pulse excitation vector output from the pulse position searcher 2105 to the multiplier 2107 is multiplied by a quantized pulse excitation vector gain quantized by an external gain quantizer, and output to the adder 2108. In the pulse position searcher 2105, the polarity and index information of each sound source pulse representing the pulse sound source vector together with the pulse sound source vector are separately output to the outside of the sound source generation unit. The polarity and index information of the excitation pulse are converted into a data sequence to be output to the transmission path through an encoder, a multiplexer, and the like, and sent out to the transmission path.

The adder 2108 performs vector addition of the adaptive code vector component output from the multiplier 2106 and the pulse excitation vector component output from the multiplier 2107, and outputs the result as an excitation excitation vector.

The index allocating method according to the present embodiment can be applied to all cases where the position information of the sound source is expressed by relative values, and the difference is only in the index allocating method. Therefore, it is possible to obtain an effect of suppressing propagation of a transmission line error without affecting performance at all.

It should be noted that the decoder side has the same index updating means as the encoder side. In addition, as a method of setting up a pulse, a constant number of pulses, for example, four pulses are searched in a search range, for example, three pulses.
When standing at any of the two positions, as described above, 32 combinations are divided into four and one pulse is determined as one of eight assigned positions, and all combinations (8 × 8 × 8 × 8 ways)
There is a method of searching for all combinations in which four locations are selected from 32 locations. In addition to the combination of the impulses having the amplitude of 1, a combination of a plurality of pulses, for example, two pulses, or a combination of impulses having different amplitudes may be used.

(Embodiment 13) FIG. 23 shows a first embodiment of the present invention.
3 shows the third embodiment, and includes a pulse number and index updating means for assigning a pulse search position index and a pulse number, and determines a pulse position search range based on a pitch period and a pitch peak position of an adaptive code vector. 2 shows a sound source generation unit on the encoder side of the CELP type speech encoding device. More specifically, in a CELP-type speech coding apparatus that performs excitation pulse search at a relative position from a pitch peak position, a pulse position is indexed sequentially from the head of a subframe, and different numbers having the same index number are used. For each pulse, the pulse numbers are assigned in order from the head of the subframe, that is, in the case of the same index number, the pulse number is determined so that the smaller the pulse number, the closer to the head of the subframe. 2 shows a sound source generation unit that prevents the influence of a transmission path error occurring in a certain frame from propagating to a subsequent frame without a transmission path error.

In FIG. 23, reference numeral 2301 denotes an adaptive codebook for storing the past excitation source vector and outputting the selected adaptive code vector to the pitch peak position calculator 2302 and pitch gain multiplier 2306. Calculates the pitch peak position using the adaptive code vector and the pitch period L output from the as input, and outputs the pitch peak position calculator 2303 to the search position calculator 2303.
303, a search position calculator that calculates a range in which to search for a pulse sound source by using the pitch peak position and the pitch period L output from the pitch peak position calculator 2302 as input, and outputs the range to the pulse number and index updating unit 2304;
A pulse number and index updating unit 2304 outputs the pulse number and index output from the search position calculator 2303 to the pulse position search unit 2305 by changing the number of each sound source pulse and the index of each position of each sound source pulse. A pulse sound source is input using the search position (the pulse number and the index indicating the pulse position are output again) output from the index updating unit 2304 and the pitch period L separately calculated outside the sound source generation unit. A pulse position searcher that searches for and outputs a pulse excitation vector to a pulse excitation gain multiplier 2307 and outputs an index representing the pulse excitation vector to the outside of the excitation generator as an encoded output. Multiplied by the adaptive code vector gain to the adaptive code vector Multiplier outputs 08, 2307 adder multiplies a pulse excitation vector gain pulse excitation vector outputted from the pulse position searcher 2305 230
8 is an adder which receives the output from the multiplier 2306 and the output from the multiplier 2307 as inputs, adds the vectors, and outputs the result as an excitation sound source vector.

The operation of the sound source generator configured as described above will be described with reference to FIGS. In FIG. 23, adaptive codebook 2301 cuts out an adaptive code vector by a subframe length from a point that has been traced back by a pitch period L calculated in advance outside the excitation generation unit, and outputs the cutout as an adaptive code vector. If the pitch period L is less than the subframe length, a vector obtained by repeatedly connecting the extracted pitch period L until the subframe length is reached is output as an adaptive code vector.

The pitch peak position calculator 2302 determines the position of the pitch peak existing in the adaptive code vector using the adaptive code vector output from the adaptive code book 2301. The position of the pitch peak can be determined by maximizing the normalized cross-correlation between the impulse train arranged in the pitch cycle and the adaptive code vector. Further, it is possible to obtain the impulse trains arranged in the pitch cycle through a synthesis filter, and to minimize the error caused by passing the adaptive code vector through the synthesis filter, thereby obtaining the error more accurately.

The search position calculator 2303 determines the search position of the excitation pulse based on the pitch peak position, and outputs it to the pulse number and index updating means 2304. As a method of determining the search position, for example, as described in the sixth and eighth embodiments, the search position is determined so that the vicinity of the pitch peak is densely distributed and the other portions are sparsely distributed. Note that, as described in Embodiments 6 and 8, the number of sound source pulses is changed using the pitch period information,
It is also effective to limit the search range of the sound source pulse. Examples of specific search positions determined by the search position calculator 2303 are shown in FIGS.
(B), FIG. 11 (c), and FIG. For example, FIG. 10 specifically shows a method of limiting the pulse position search range densely in the vicinity of the pitch pulse position and sparsely in other portions. This limiting method is based on a statistical result in which positions where the pulse is likely to be raised are concentrated near the pitch pulse. If the pulse position search range is not limited, the probability that a pulse will be raised near the pitch pulse in a voiced part will be higher than the probability that it will be raised in other parts. In addition, what is calculated by the search position calculator is
This is the search position of the sound source pulse using the relative position from the pitch peak position. At this point, the pulse number and index are assigned in ascending order of the numerical value of the relative position with the pitch peak position being 0 (see FIG. 24 (b)
reference). FIG. 24 shows a case corresponding to FIGS. 11B and 13 when the number of pulses is four. FIG. 24A shows the sound source pulse search positions determined by the search position calculator 2103 when the number of pulses is four, and the length of the arrow and the upward and downward directions indicate the four types of sound source pulse search positions. Is shown. In the relative position of FIG. 24A, each sample point is represented by a numerical value from -4 to +75, with the pitch peak position being 0, and points before -4 are turned back to points protruding beyond the subframe boundary. This is represented by the numerical value of +.

The pulse number and index updating means 2304 is indexed in ascending order of relative position from the pitch peak position (FIG. 24).
(B)) The sound source pulse search position is converted to an absolute position where the head of the subframe is 0, and then the pulse number and index are re-assigned in ascending order of the absolute position (FIG. 24 (c)). Output to searcher 2305. In this way, when the pitch peak positions calculated on the encoder side and the decoder side are different due to a transmission path error or the like, the pulse position shift can be reduced.

The pulse position searcher 2305 uses the pulse period L input separately from the sound source pulse search position at which the index indicating the search position has been re-assigned by the pulse number and index updating means 2304, and Determine the optimal combination of pulse positions.
For the method of pulse search, see `` ITU-T Recommendation G.729: Cod
ing of Speech at 8 kbits / s using Conjugate-Structu
re Algebraic-Code-Excited Linear-Prediction (CS-AC
ELP), March 1996, for example, when the number of pulses is four, the combination of i0 to i3 is determined so as to maximize the equation (2) shown in the sixth embodiment. Note that the polarity of each excitation pulse at this time is the target vector of the noise codebook component, that is, a signal obtained by subtracting the zero input response signal of the auditory weighting synthesis filter and the signal of the adaptive codebook component from the auditory weighted input speech. If the pulse position is determined in advance before performing the pulse position search so as to be equal to the polarity at each position of the vector, the calculation amount for the search can be greatly reduced. When the pitch period is shorter than the subframe length, a pitch period filter is applied as described in the fifth embodiment, so that the excitation pulse is not an impulse but a pulse train of the pitch period. In the case of performing such pitch periodization processing, if a pitch periodization filter is previously applied to the impulse response vector of the auditory weighting synthesis filter, the maximum value of Expression (2) can be obtained in the same manner as when pitch periodization is not performed. The search for the sound source pulse can be performed by the conversion. A pulse is generated at the position of each sound source pulse determined in this way according to the polarity of each determined sound source pulse, and if a pitch period filter is applied using the pitch period L, a pulse sound source vector is generated. The generated pulse sound source vector is
07. The pulse excitation vector output from the pulse position searcher 2305 to the multiplier 2307 is multiplied by a quantized pulse excitation vector gain quantized by an external gain quantizer, and output to the adder 2308. In the pulse position searcher 2305, the polarity and index information of each excitation pulse representing the pulse excitation vector together with the pulse excitation vector are separately output outside the excitation generator. The polarity and index information of the excitation pulse are converted into a data sequence to be output to the transmission path through an encoder, a multiplexer, and the like, and sent out to the transmission path.

The adder 2308 performs vector addition of the adaptive code vector component output from the multiplier 2306 and the pulse excitation vector component output from the multiplier 2307, and outputs the result as an excitation excitation vector.

The index assignment method according to the present embodiment can be applied to all cases where the position information of the sound source is expressed by relative values. Therefore, the effect of suppressing propagation of transmission path errors can be obtained without affecting performance. Further, by switching and using the pulse sound source at the fixed search position, the propagation of the influence of the transmission path error can be further suppressed.

The decoder side also has a similar pulse number and index updating means 2304. In addition, as a method of setting a pulse, when a fixed number of, for example, four pulses are set in a search range, for example, somewhere in 32 positions, as described above, 32 positions are divided into four and one pulse is set. In addition to the method of searching all combinations (8 × 8 × 8 × 8 ways) so as to determine one of eight locations to which is assigned, all combinations that select four locations from 32 locations There are ways to search. In addition to the combination of the impulses of the amplitude 1, a plurality
It is also possible to form a pulse by a combination of a pulse pair obtained by combining these pulses or a combination of impulses having different amplitudes.

(Embodiment 14) FIG. 25 shows a first embodiment of the present invention.
7 shows the sound source generation unit of the CELP-type speech coding apparatus that performs a pulse search using the sound source pulse search positions generated by both the fixed search position and the phase adaptive search position.

In FIG. 25, reference numeral 2501 denotes an adaptive codebook for storing the past excitation source vector and outputting the selected adaptive code vector to the pitch peak position calculator 2502 and pitch gain multiplier 2506. Reference numeral 2502 denotes the adaptive codebook 2501. A pitch peak position calculator that calculates a pitch peak position by using the adaptive code vector output from the input device and a pitch period L input from the outside as inputs, and outputs a pitch peak position calculator 2503 to the search position calculator 2503.
A search position calculator that calculates a pulse sound source search position using the pitch peak position output from 502 and the pitch period L input from the outside as inputs, and outputs the calculated position to an adder 2504, which outputs from the search position calculator 2503. The search position represented by the relative position with the pitch peak position being 0 and the search position searched at the fixed position are combined (not a numerical addition, but a sum of a set of two types of search positions. The adder outputs to the pulse position searcher 2505. The search position 2505 searches for a pulse sound source using the search position output from the adder 2504 and the pitch period L separately calculated outside the sound source generator as inputs. A pulse position searcher that outputs an excitation vector to a pulse excitation gain multiplier 2507, and an adaptive code vector 2506 output from an adaptive codebook 2501 Is multiplied by an adaptive code vector gain and output to an adder 2508. A multiplier 2507 multiplies the pulse excitation vector output from the pulse position searcher 2505 by a pulse excitation vector gain and outputs the result to an adder 2508. An adder that receives an output from the multiplier 2506 and an output from the multiplier 2507 as inputs, performs vector addition, and outputs the resultant as an excitation sound source vector.

The operation of the sound source generator configured as described above will be described with reference to FIGS. 25 and 26. In FIG. 25, adaptive codebook 2501 cuts out an adaptive code vector by a subframe length from a point that has been advanced in the past by a pitch period L calculated in advance outside the excitation generation unit, and outputs it as an adaptive code vector. If the pitch period L is less than the subframe length, a vector obtained by repeatedly connecting the extracted pitch period L until the subframe length is reached is output as an adaptive code vector.

The pitch peak position calculator 2502 uses the adaptive code vector output from the adaptive code book 2501 to determine the position of the pitch peak existing in the adaptive code vector. The position of the pitch peak can be determined by maximizing the normalized cross-correlation between the impulse train arranged in the pitch cycle and the adaptive code vector. Further, an error in which the impulse train arranged in the pitch cycle is passed through the synthesis filter and an original error in passing the adaptive code vector through the synthesis filter are minimized (maximize the normalized cross-correlation function).
By doing so, it is also possible to obtain it with higher accuracy.

The search position calculator 2503 determines the search position of the excitation pulse based on the pitch peak position, and outputs it to the adder 2504. As a method of determining the search position, for example, as shown in FIG. 26, a method of outputting a point that does not overlap with the fixed search position near the pitch peak is used. The same applies to the case where the number of sound source pulses is changed using the pitch period information or the search range of the sound source pulse is limited as described in the sixth and eighth embodiments. Examples of specific search positions determined by the search position calculator 2503 are shown in FIGS. In FIG. 26, the fixed search position is set to an odd-numbered sample point (FIG. 26A), and the search position calculator 2503 sets the search position to an even-numbered sample point near the pitch peak (FIGS. 26B and 26B). c)). FIG.
6B shows a case where the pitch peak position is at an even-numbered sample point (the pitch peak position is not included in the fixed search position), and FIG. 26C shows a case where the pitch peak position is at an odd-numbered sample point (when the pitch peak position is (Included in the fixed search position)
Each case is shown. As can be seen from the comparison between FIGS. 26B and 26C, the search position (the relative position where the pitch peak position is 0) slightly differs depending on the position of the pitch peak position.

The adder 2504 is provided with a search position calculator 250
A set of sound source pulse search positions output from FIG.
(B), (c)) and a predetermined set of fixed search positions (FIG. 26 (a)) are obtained and output to the pulse position searcher 2505 (FIG. 26 (d)). By doing so, the search position of the sound source pulse is limited densely in the vicinity of the pitch peak position and sparsely in other portions. This limiting method is based on a statistical result in which positions where the pulse is likely to be raised are concentrated near the pitch pulse. If the pulse position search range is not limited, the probability that a pulse will be raised near the pitch pulse in a voiced part will be higher than the probability that it will be raised in other parts. If the calculation of the pitch peak position on the decoder side is erroneous due to the influence of a transmission path error or the like, the search position of the excitation pulse calculated by the search position calculator 2503 differs between the encoder side and the decoder side. Since a part of the excitation pulse search position input to the pulse position search unit 2505 is a fixed search position, it is possible to reduce the probability that the pulse positions of the encoder side and the decoder side differ from each other. The effect of a road error can be reduced.

[0187] The pulse position search unit 2505 includes the adder 25.
Using the sound source pulse search position output from the signal generator 04 and the pitch period L separately input, an optimal combination of the positions where the sound source pulses are set is determined. The method of pulse search is "IT
UT Recommendation G.729: Coding of Speech at 8 kbi
ts / s using Conjugate-Structure Algebraic-Code-Exci
As shown in “ted Linear-Prediction (CS-ACELP), March 1996”, for example, when the number of pulses is four, i0 to i3 are used to maximize the equation (2) shown in the sixth embodiment. Is determined. Note that the polarity of each excitation pulse at this time is the target vector of the noise codebook component, that is, a signal obtained by subtracting the zero input response signal of the auditory weighting synthesis filter and the signal of the adaptive codebook component from the auditory weighted input speech. If the pulse position is determined in advance before performing the pulse position search so as to be equal to the polarity at each position of the vector, the calculation amount for the search can be greatly reduced. When the pitch period is shorter than the subframe length, a pitch period filter is applied as described in the fifth embodiment, so that the excitation pulse is not an impulse but a pulse train of the pitch period. In the case of performing such pitch periodization processing, if a pitch periodization filter is previously applied to the impulse response vector of the auditory weighting synthesis filter, the maximum value of Expression (2) can be obtained in the same manner as when pitch periodization is not performed. The search for the sound source pulse can be performed by the conversion. A pulse is generated at the position of each sound source pulse determined in this way according to the polarity of each determined sound source pulse, and if a pitch period filter is applied using the pitch period L, a pulse sound source vector is generated. The generated pulse excitation vector is output to multiplier 2507.
The pulse excitation vector output from the pulse position searcher 2505 to the multiplier 2507 is multiplied by the quantized pulse excitation vector gain quantized by the external gain quantizer, and output to the adder 2508. Although omitted in FIG. 25, in the pulse position searcher 2505,
The polarity and index information of each excitation pulse representing the pulse excitation vector together with the pulse excitation vector are separately output to the outside of the excitation generator. The polarity and index information of the excitation pulse are converted into a data sequence to be output to the transmission path through an encoder, a multiplexer, and the like, and sent out to the transmission path.

Adder 2508 performs vector addition of the adaptive code vector component output from multiplier 2506 and the pulse excitation vector component output from multiplier 2507, and outputs the result as an excitation excitation vector.

[0189] It is possible to further suppress the propagation of the effect of the transmission path error by switching and using the pulse sound source at the fixed search position.

As for the method of setting the pulses, when a fixed number of pulses, for example, four pulses are set in a search range, for example, somewhere in 32 positions, the 32 positions are divided into four as described above, and one pulse is set. All of the combinations (8 × 8
In addition to the method of searching (× 8 × 8 ways), there is a method of searching for all the combinations that select four places from 32 places. In addition to the combination of the impulses having the amplitude of 1, a combination of a plurality of pulses, for example, two pulses, or a combination of impulses having different amplitudes may be used.

(Embodiment 15) FIG. 27 shows a first embodiment of the present invention.
15 shows an excitation generator of the CELP-type speech coding apparatus according to the fifth embodiment, which is provided with a pitch peak position corrector.

In FIG. 27, reference numeral 2701 denotes an adaptive codebook which stores the past excitation source vector and outputs the selected adaptive code vector to pitch peak position calculator 2702, pitch peak position corrector 2703, and pitch gain multiplier 2706. , 2702 calculate the pitch peak position using the adaptive code vector output from the adaptive codebook 2701 and the externally input pitch period L as inputs, and output a pitch peak position calculator to a pitch peak position corrector 2703. Adaptive code vector output from adaptive codebook 2701 and pitch peak position calculator 2702
The pitch peak position output from and the pitch period L input from the outside are input to correct the pitch peak position,
Pitch peak position corrector output to search position calculator 2704. 2704 receives as input the pitch peak position output from pitch peak position corrector 2703 and separately input pitch period L, and sets the search position of the sound source pulse to the pulse position. Search position calculator for outputting to searcher 2705, 2705
Is the search position output from the search position calculator 2704,
A pulse position searcher that searches for a pulse excitation using the pitch period L separately calculated outside the excitation generation unit as an input and outputs a pulse excitation vector to a pulse excitation gain multiplier 2707 is output from the adaptive codebook 2701. The multiplier 2707 multiplies the adaptive code vector by the adaptive code vector gain and outputs the result to the adder 2708. The multiplier 2707 multiplies the pulse excitation vector output from the pulse position searcher 2705 by the pulse excitation vector gain and outputs the result to the adder 2708. And 2708 are the output from the multiplier 2706 and the multiplier 2706.
This is an adder that receives the output from 707 as an input, adds the vector, and outputs the result as an excitation sound source vector.

The operation of the sound source generator configured as described above will be described with reference to FIGS. 27 and 28. In FIG. 27, adaptive codebook 2701 cuts out an adaptive code vector by a subframe length from a point that has been advanced in the past by a pitch period L calculated in advance outside the excitation generation unit, and outputs it as an adaptive code vector. If the pitch period L is less than the subframe length, a vector obtained by repeatedly connecting the extracted pitch period L until the subframe length is reached is output as an adaptive code vector.

Pitch peak position calculator 2702 uses the adaptive code vector output from adaptive code book 2701 to determine the position of the pitch peak existing in the adaptive code vector. The position of the pitch peak can be determined by maximizing the normalized cross-correlation between the impulse train arranged in the pitch cycle and the adaptive code vector. Further, an error in which the impulse train arranged in the pitch cycle is passed through the synthesis filter and an original error in passing the adaptive code vector through the synthesis filter are minimized (maximize the normalized cross-correlation function).
By doing so, it is also possible to obtain it with higher accuracy.

Pitch peak position corrector 2703 calculates the adaptive code vector output from adaptive codebook 2701
A vector having a length of one pitch cycle length L including the point of the pitch peak position calculated by the pitch peak position calculator 2702 is cut out, and a point having the maximum amplitude value is searched out from the cut out waveform to search for a position. Calculator 270
4 is output. This process is performed only when the pitch period L is shorter than the subframe length. If the pitch period L is longer than the subframe length, the pitch peak position output from the pitch peak position calculator 2702 is output to the pulse position searcher 2705 as it is. The pitch peak position output from the pitch peak position calculator 2702 may be the second highest amplitude place in one pitch waveform when one subframe length is equivalent to about one pitch period. (FIGS. 28A and 28B: There is only one pitch peak in one subframe,
(Since there are two points (second peaks) having the second largest amplitude value in one pitch period waveform in one subframe, the second peak is erroneously detected as a pitch peak.) Therefore, the pitch peak position corrector 2703
By checking whether there is a point having a larger amplitude value within one pitch cycle length from the pitch peak position output from the pitch peak position calculator 2702, the vicinity of the pitch peak position output from the pitch peak position calculator 2702 is checked. When there is a point having an amplitude value larger than the amplitude value of the point, the point having the larger amplitude value is set as the pitch peak position. For example, when the pitch peak position calculator 2702 outputs the second peak in FIG. 28 (c), the amplitude is the largest in the adaptive code vector for one pitch period from the second peak (the thick line portion in FIG. 28 (c)). Is a pitch peak.

The search position calculator 2704 determines the search position of the excitation pulse based on the pitch peak position output from the pitch peak position corrector 2703, and outputs it to the pulse position searcher 2705. As a method of determining the search position, the search position of the sound source pulse is limited densely in the vicinity of the pitch peak position and sparsely in other portions, as in the fifth embodiment, the sixth embodiment or the fourteenth embodiment. There is a way to do that. This limiting method is based on a statistical result in which positions where the pulse is likely to be raised are concentrated near the pitch pulse. In the case where the pulse position search range is not limited, the fact that the probability that a pulse is raised near a pitch pulse in a voiced part is higher than the probability that a pulse is raised in other parts is used.

The pulse position search unit 2705 determines the optimum combination of the position where the sound source pulse is to be formed using the sound source pulse search position output from the search position calculator 2704 and the pitch period L separately input. The method of pulse search is described in "ITU-T Recommendation G.729: Coding of Speecha
t 8 kbits / s using Conjugate-Structure Algebraic-Co
de-Excited Linear-Prediction (CS-ACELP), March 199
As shown in “6”, for example, when the number of pulses is four, the combination of i0 to i3 is determined so as to maximize Expression (2) shown in the sixth embodiment. Note that the polarity of each excitation pulse at this time is the target vector of the noise codebook component, that is, a signal obtained by subtracting the zero input response signal of the auditory weighting synthesis filter and the signal of the adaptive codebook component from the auditory weighted input speech. If the pulse position is determined in advance before performing the pulse position search so as to be equal to the polarity at each position of the vector, the calculation amount for the search can be greatly reduced. When the pitch period is shorter than the subframe length, the fifth embodiment
By applying a pitch period filter as described above, the sound source pulse is not an impulse but a pulse train having a pitch period. In the case of performing such pitch periodization processing, if a pitch periodization filter is previously applied to the impulse response vector of the auditory weighting synthesis filter, the maximum value of Expression (2) can be obtained in the same manner as when pitch periodization is not performed. The search for the sound source pulse can be performed by the conversion. A pulse is generated at the position of each sound source pulse determined in this way according to the determined polarity of each sound source pulse, and a pitch sound source vector is generated by applying a pitch period filter using the pitch period L. The generated pulse excitation vector is output to multiplier 2707. The pulse excitation vector output from the pulse position searcher 2705 to the multiplier 2707 is multiplied by a quantized pulse excitation vector gain quantized by an external gain quantizer and output to the adder 2708. Although omitted in FIG. 27, in the pulse position searcher 2705 of the encoder, the polarity and index information of each excitation pulse representing the pulse excitation vector together with the pulse excitation vector are separately output outside the excitation generation unit. The polarity and index information of the excitation pulse are converted into a data sequence to be output to the transmission path through an encoder, a multiplexer, and the like, and sent out to the transmission path.

Adder 2708 performs vector addition of the adaptive code vector component output from multiplier 2706 and the pulse excitation vector component output from multiplier 2707, and outputs the resultant as an excitation excitation vector.

In this embodiment, the index updating means or the pulse number and index updating means or the fixed search position and the phase adaptive search position are different from those in the twelfth, thirteenth, or fourteenth embodiment. By adopting the combination of the above, the influence of the transmission path error can be reduced. Further, by switching and using the pulse sound source at the fixed search position, the propagation of the influence of the transmission path error can be further suppressed.

Further, the pitch peak position corrector of the present invention can be applied to any of the speech coding apparatuses according to the third to eleventh embodiments.

As for the method of setting pulses, when a constant number of pulses, for example, four pulses are set in a search range, for example, somewhere in 32 positions, 32 positions are divided into four as described above, and one pulse is set. All of the combinations (8 × 8
In addition to the method of searching (× 8 × 8 ways), there is a method of searching for all the combinations that select four places from 32 places. In addition to the combination of the impulses having the amplitude of 1, a combination of a plurality of pulses, for example, two pulses, or a combination of impulses having different amplitudes may be used.

(Embodiment 16) FIG. 29 shows the first embodiment of the present invention.
6 shows an embodiment 6 in which the existence range of the pitch peak position is previously limited before calculating the pitch peak position by using the continuity of the phase of the sound source signal waveform between consecutive subframes.
2 shows a sound source generation unit of the ELP type speech coding apparatus. In FIG. 29, reference numeral 2901 denotes an adaptive codebook that outputs an adaptive code vector to a pitch peak position calculator 2902 and a multiplier 2908, and reference numeral 2902 denotes an adaptive code vector output from the adaptive codebook 2901 and input from outside the speech generation unit. Taking the pitch period L and the pitch peak search range output from the pitch peak search range limiter 2903 as input, calculate the pitch peak position in the adaptive code vector, and
And a search position calculator 2906 that outputs a pitch peak position calculator 2903, and 2903 is a pitch peak position in the immediately preceding subframe output from the delay unit 2904 and the delay unit 2906.
Using the pitch period in the immediately preceding subframe output from 905 and the pitch period L in the current subframe input from outside the sound source generation unit as input, predicts the pitch peak position in the current subframe, and predicts the predicted pitch. A pitch peak search range limiter that limits the range in which the pitch peak position is searched based on the peak position and outputs the range to the pitch peak position calculator 2902, and the delay unit 2904 outputs the pitch output from the pitch peak position calculator. A delay device that receives a peak position as input and delays it by one subframe and outputs it to a pitch peak search range limiter 2903. A delay unit 2905 receives a pitch period L input from outside the voice generation unit as input and delays it by one subframe. The delay device that outputs to the pitch peak search range limiter 2903 A search position calculator that outputs a search position of a sound source pulse to a pulse position searcher 2907 by using the pitch peak position output from the peak position calculator 2902 and the pitch period L input from outside the sound source generator as inputs. Receives the search position of the sound source pulse input from the search position calculator 2906 and the pitch period L input from outside the sound source generator, and uses the input sound source pulse search position and the pitch period L to generate the sound source pulse. A pulse position searcher that searches for a position and outputs a pulse excitation vector to a multiplier 2909, receives the adaptive code vector output from the adaptive codebook as an input, multiplies the quantized adaptive code vector gain, and outputs the result to an adder 2910. Multiplier, 29
09 is a multiplier which receives the pulse excitation vector output from the pulse position searcher 2907 as input, multiplies the quantization pulse excitation vector gain, and outputs the result to the adder 2910;
Is an adder that receives the vectors output from the multipliers 2908 and 2909 as inputs, performs addition of the input vectors, and outputs the result as an excitation sound source vector.

[0203] The operation of the sound source generation unit of the speech coding apparatus configured as described above will be described with reference to FIG. The adaptive codebook 2901 is constituted by a buffer of a past excitation source, extracts a corresponding portion from the buffer of the excitation source based on a pitch period or pitch lag obtained by an external pitch analysis or an adaptive codebook search means, and generates an adaptive codebook. The vector is output to the pitch peak position calculator 2902 and the multiplier 2908 as a vector. The adaptive code vector output from the adaptive codebook 2901 to the multiplier 2908 is multiplied by a quantized adaptive code vector gain quantized by an external gain quantizer and added to the adder 2910.
Is output to

[0204] Pitch peak position calculator 2902 detects a pitch peak from the adaptive code vector, and outputs the position to delay device 2904 and search position calculator 2906, respectively. The detection (calculation) of the pitch peak position can be performed by maximizing the normalized cross-correlation function between the impulse train vector arranged in the pitch period L and the adaptive code vector. Also, maximizing a normalized cross-correlation function between a vector obtained by convolving the impulse response of the synthesis filter with the impulse train vector arranged in the pitch period L and a vector obtained by convolving the impulse response of the synthesis filter with the adaptive code vector. Accordingly, the pitch peak position can be detected with higher accuracy. Furthermore, if post-processing is performed to set the position where the amplitude value becomes maximum among the one-pitch periodic waveforms including the detected pitch-peak position as the pitch peak, erroneous detection of the second peak in the one-pitch periodic waveform is avoided. It is also possible.

The delay unit 2904 delays the pitch peak position calculated by the pitch peak position calculator 2902 by one subframe, and sets the pitch peak search range limiter 29
03 is output. That is, the pitch peak search range limiter 2
To 903, the pitch peak position in the immediately preceding subframe is input from the delay unit 2904. Delay unit 2905
Sets the pitch period L input from outside the sound source generation unit to 1
Delayed by a subframe and output to pitch peak search range limiter 2903. That is, the pitch period in the immediately preceding subframe is input from the delay unit 2905 to the pitch peak search range limiter 2903.

The pitch peak search range limiter 2903 is
First, a comparison is made between the pitch cycle in the immediately preceding subframe input from the delay unit 2905 and the pitch cycle in the current subframe, and it is determined whether the current subframe is a voiced (stationary) part. In particular,
If the difference between the pitch cycle in the immediately preceding subframe and the pitch cycle in the current subframe is small (for example, within ± 5 samples), it is determined to be a voiced (stationary) part. Note that voiced determination can be performed using a pitch period up to several subframes before the number of delay units is increased. If it is determined that the voice frame is a voiced (stationary) part, the pitch peak search range limiter 2903 determines the pitch peak position in the subframe immediately before input from the delay unit 2904 and the subframe immediately before input from the delay unit 2905. , And the pitch period L in the current subframe are input, the pitch peak position in the current subframe is predicted, and the range before and after the predicted position (for example, 10 samples) is set as the range in which to search for the pitch peak position. .
When the predicted pitch peak position is near the head of the subframe, the vicinity after one pitch cycle is added to the search range, and the predicted pitch peak position is near the position one pitch cycle after the head of the subframe. In this case, the vicinity of the head of the subframe is also added to the search range. If it is determined that the pitch is not a voiced (stationary) part, the entire subframe is set as the pitch peak search range without limiting the pitch peak search range. The pitch peak search range obtained by the pitch peak search range limiter 2903 in this manner is
It is output to pitch peak position calculator 2902. In addition,
The point when the audio encoding process started (first subframe)
Since there is no pitch period L input in the past (in the immediately preceding subframe), the delay unit 2905 outputs an appropriate constant (for example, a pitch period that is impossible such as the maximum value or the minimum value of the pitch period or 0). Keep it. The same applies to the delay unit 2904. Note that the predicted pitch peak position is obtained by Expression (6) shown in Embodiment 10 (see FIG. 19).

The search position calculator 2906 determines the search position of the sound source pulse based on the pitch peak position, and outputs the search position to the pulse position searcher 2907. As a method of determining a search position, for example, the sixth embodiment or the eighth embodiment
As shown in (1), the search position is determined so that the search position is distributed densely in the vicinity of the pitch peak and sparsely in other portions. It is also effective to change the number of sound source pulses using the pitch period information or limit the search range of the sound source pulses as described in the sixth and eighth embodiments. Further, if the search position is determined as described in any of Embodiments 12 to 14, it is also possible to reduce the influence of transmission path errors.

The pulse position search unit 2907 generates a sound source pulse using the pitch period L input separately from the sound source pulse search position determined by the search position calculator 2906 or a predetermined fixed search position. To determine the optimal combination of See ITU-T Reco
mmendation G.729: Coding of Speech at 8 kbits / susi
ng Conjugate-Structure Algebraic-Code-Excited Line
ar-Prediction (CS-ACELP), March 1996 ”, for example, when the number of pulses is four, the values of i0 to i3 are set so as to maximize the expression (2) shown in the sixth embodiment. Determine the combination. Note that the polarity of each excitation pulse at this time is the target vector of the noise codebook component, that is, a signal obtained by subtracting the zero input response signal of the auditory weighting synthesis filter and the signal of the adaptive codebook component from the auditory weighted input speech. Is determined before performing the pulse position search so as to be equal in polarity at each position of the vector. When the pitch period is shorter than the subframe length, a pitch period filter is applied as described in the fifth embodiment, so that the excitation pulse is not an impulse but a pulse train of the pitch period. In the case of performing such pitch periodization processing, if a pitch periodization filter is previously applied to the impulse response vector of the auditory weighting synthesis filter, the maximum value of Expression (2) can be obtained in the same manner as when pitch periodization is not performed. The search for the sound source pulse can be performed by the conversion. A pulse is generated at the position of each sound source pulse determined in this way according to the polarity of each determined sound source pulse, and if a pitch period filter is applied using the pitch period L, a pulse sound source vector is generated. The generated pulse excitation vector is output to multiplier 2909. The pulse excitation vector output from the pulse position searcher 2907 to the multiplier 2909 is multiplied by a quantized pulse excitation vector gain quantized by an external gain quantizer and output to the adder 2910.

The adder 2910 performs vector addition of the adaptive code vector component output from the multiplier 2908 and the pulse excitation vector component output from the multiplier 2909, and outputs the result as an excitation excitation vector.

As for the method of setting pulses, when a constant number of pulses, for example, 4 pulses are set in a search range, for example, somewhere in 32 positions, 32 positions are divided into four as described above, and 1 pulse is set. All combinations (8 × 8) are determined so that one pulse is determined to be one of eight assigned pulse positions.
In addition to the method of searching (× 8 × 8 ways), there is a method of searching for all the combinations that select four places from 32 places. In addition to the combination of the impulses having the amplitude of 1, a combination of a plurality of pulses, for example, two pulses, or a combination of impulses having different amplitudes may be used.

(Embodiment 17) FIG. 30 shows a first embodiment of the present invention.
7 shows a seventh embodiment in which a pulse searcher uses a fixed search position in which the number of pulses is small and the position information assigned to each pulse is sufficient, and a position in which the number of pulses is large and assigned to each pulse. A CELP-type speech comprising a pulse searcher using a sound source pulse search position where information is not always sufficient, and a selector for selecting an optimum pulse sound source vector from among the pulse sound source vectors output from the plurality of pulse searchers 3 shows a sound source generation unit of the encoding device.

In FIG. 30, reference numeral 3001 denotes an adaptive codebook for storing a past excitation source vector and outputting a selected adaptive code vector to a pitch peak position calculator 3002 and a pitch gain multiplier 3007; Calculates the pitch peak position by using the adaptive code vector output from the PID and the pitch period L input from the outside, and outputs the pitch peak position to the search position calculator 3003.
A search position calculator that outputs a search position of a sound source pulse to a pulse position searcher 3004 by using the pitch peak position output from 002 and the pitch period L input from outside the sound source generation unit as inputs, and 3004 calculates a search position. Container 3003
A pulse position searcher that searches for a pulsed sound source by using the search position output from the above and a pitch period L separately calculated outside the sound source generation unit and outputs a pulsed sound source vector 1 to a selector 3005, Position searcher 300
The selector 3006 receives the pulse excitation vector 1 output from the pulse excitation vector 4 and the pulse excitation vector 2 output from the pulse position searcher 3006, selects the optimal pulse excitation vector, and outputs the selected pulse excitation vector to the multiplier 3008. A pulse position searcher that searches for a pulse sound source by using a fixed fixed search position and a pitch period L input from the outside of the sound source generation unit and outputs it as a pulse sound source vector 2 to a selector 3005. Reference numeral 3007 denotes an adaptive codebook. A multiplier that multiplies the adaptive code vector output from 3001 by the adaptive code vector gain and outputs the result to the adder 3009;
A multiplier that multiplies the pulse excitation vector output from 005 by the pulse excitation vector gain and outputs the result to the adder 3009. The output 3009 is the output from the multiplier 3007 and the multiplier 309.
This is an adder that receives an output from the input device 08, adds a vector, and outputs the resultant as an excitation sound source vector.

The operation of the sound source generator configured as described above will be described with reference to FIG. In FIG. 30, adaptive codebook 3001 cuts out an adaptive code vector by a subframe length from a point that has been retroactive by a pitch period L calculated in advance outside the excitation generation unit, and outputs the cutout as an adaptive code vector. If the pitch period L is less than the subframe length, a vector obtained by repeatedly connecting the extracted pitch period L until the subframe length is reached is output as an adaptive code vector.

[0214] Pitch peak position calculator 3002 determines the position of the pitch peak existing in the adaptive code vector using the adaptive code vector output from adaptive code book 3001. The position of the pitch peak can be determined by maximizing the normalized cross-correlation between the impulse train arranged in the pitch cycle and the adaptive code vector. Further, an error in which the impulse train arranged in the pitch cycle is passed through the synthesis filter and an original error in passing the adaptive code vector through the synthesis filter are minimized (maximize the normalized cross-correlation function).
By doing so, it is also possible to obtain it with higher accuracy.
In addition, if the pitch peak corrector as shown in the fifteenth embodiment is provided, it is possible to reduce the calculation error of the pitch peak position.

The search position calculator 3003 determines the search position of the excitation pulse based on the pitch peak position output from the pitch peak position calculator 3002, and outputs it to the pulse position searcher 3004. As a method of determining the search position, the search position of the sound source pulse is limited densely in the vicinity of the pitch peak position and sparsely in other portions, as in the fifth embodiment, the sixth embodiment or the fourteenth embodiment. There is a way to do that. This limiting method is based on a statistical result in which positions where the pulse is likely to be raised are concentrated near the pitch pulse. In the case where the pulse position search range is not limited, the fact that the probability that a pulse is raised near a pitch pulse in a voiced part is higher than the probability that a pulse is raised in other parts is used. Note that the use of a method for determining a sound source pulse search position as described in any of Embodiments 12 to 14 can also reduce the effect of transmission path errors.

The pulse position searcher 3004 determines the optimum combination of the position where the sound source pulse is to be formed using the sound source pulse search position output from the search position calculator 3003 and the pitch period L separately input. The method of pulse search is described in "ITU-T Recommendation G.729: Coding of Speecha
t 8 kbits / s using Conjugate-Structure Algebraic-Co
de-Excited Linear-Prediction (CS-ACELP), March 199
As shown in “6”, for example, when the number of pulses is four, the combination of i0 to i3 is determined so as to maximize Expression (2) shown in the sixth embodiment. Note that the polarity of each excitation pulse at this time is the target vector of the noise codebook component, that is, a signal obtained by subtracting the zero input response signal of the auditory weighting synthesis filter and the signal of the adaptive codebook component from the auditory weighted input speech. If the pulse position is determined in advance before performing the pulse position search so as to be equal to the polarity at each position of the vector, the calculation amount for the search can be greatly reduced. When the pitch period is shorter than the subframe length, the fifth embodiment
By applying a pitch period filter as described above, the sound source pulse is not an impulse but a pulse train having a pitch period. In the case of performing such pitch periodization processing, if a pitch periodization filter is previously applied to the impulse response vector of the auditory weighting synthesis filter, the maximum value of Expression (2) can be obtained in the same manner as when pitch periodization is not performed. The search for the sound source pulse can be performed by the conversion. A pulse is generated at the position of each sound source pulse determined in this way according to the determined polarity of each sound source pulse, and a pitch sound source vector is generated by applying a pitch period filter using the pitch period L. The generated pulse excitation vector is output to the selector 3005 as the pulse excitation vector 1. The pulse position searcher 300
The sound source pulse search position used in No. 4 has a large number of sound source pulses, so that the position information allocated to each sound source pulse is not always sufficient. That is, the mode in which the pulse position searcher 3004 is used is a mode in which the number of pulses is large but the position of each pulse cannot always be strictly represented. When the position information of each pulse is insufficient, the effect of using the method of determining the pulse search position as performed by the search position calculator 3003 can be obtained.

The pulse position searcher 3006 determines an optimum combination of a position where a sound source pulse is to be set, using a predetermined fixed search position and a pitch period L separately input from outside the sound source generation unit. The method of pulse search is “ITU-
T Recommendation G.729: Coding of Speech at 8 kbit
s / s using Conjugate-Structure Algebraic-Code-Excit
As shown in “ed Linear-Prediction (CS-ACELP), March 1996”, for example, when the number of pulses is four, i0 to i are set so as to maximize the equation (2) shown in the sixth embodiment.
3 is determined. Note that the polarity of each excitation pulse at this time is the target vector of the noise codebook component, that is, a signal obtained by subtracting the zero input response signal of the auditory weighting synthesis filter and the signal of the adaptive codebook component from the auditory weighted input speech. If the pulse position is determined in advance before performing the pulse position search so as to be equal to the polarity at each position of the vector, the calculation amount for the search can be greatly reduced. When the pitch period is shorter than the subframe length, a pitch period filter is applied as described in the fifth embodiment, so that the excitation pulse is not an impulse but a pulse train of the pitch period. In the case of performing such a pitch periodization process, if the pitch periodization filter is applied to the impulse response vector of the auditory weighting synthesis filter in advance, the maximization of Expression (2) is performed in the same manner as the case where the pitch periodization is not performed. Thus, a search for a sound source pulse can be performed. A pulse is generated at the position of each sound source pulse determined in this way according to the polarity of each determined sound source pulse, and if a pitch period filter is applied using the pitch period L, a pulse sound source vector is generated. The generated pulse excitation vector is selected as a pulse excitation vector 2 by the selector 300.
5 is output. Here, the fixed search position input to the pulse position searcher 3006 is set so that the position information assigned to each sound source pulse is sufficient (specifically, all points in the subframe are in the fixed search position pattern). Must be narrowed down to the number of source pulses. By reducing the number of pulses and accurately representing the position at which the pulse is raised, it is possible to improve the quality of synthesized speech in a voiced rising portion or the like. Further, by providing such a mode in which the positional information is sufficient, it is possible to avoid deterioration that occurs when only the mode in which the positional information is insufficient is used.

Although FIG. 30 shows two types of pulse position searchers, it is possible to increase the number of types to three or more and perform switching according to the characteristics of the input signal. Also, even if the sound source pulse search position input to the pulse position search unit 3004 is set to a predetermined fixed search position instead of the one output from the search position calculator 3003, it is assigned to each pulse. The configuration having a mode with a small number of pulses with sufficient position information avoids the effect of improving the synthesized voice quality in a voiced rising portion or the like and the deterioration of the synthesized voice quality that occurs when only the mode with insufficient position information is used. The effect is obtained. However, using the sound source pulse search position determined by the search position calculator 3003, the pulse position searcher 30
In the case of the voice part where the sound source pulse is easily formed near the pitch peak, the use efficiency of the mode having a large number of pulses can be improved when the pulse position search is performed by the signal line 04.

The selector 3005 is the pulse position searcher 30
The pulse excitation vector 1 output from the signal generator 04 and the pulse excitation vector 2 output from the pulse position searcher 3006 are compared, and the one in which the distortion of the synthesized speech is smaller is output to the multiplier 3008 as the optimal pulse excitation vector. The pulse excitation vector output from the selector 3005 to the multiplier 3008 is multiplied by a quantized pulse excitation vector gain quantized by an external gain quantizer and added to the adder 30.
09 is output. Although omitted in FIG. 30, the pulse position searchers 3004 and 3006 of the encoder separately output the polarity and index information of each excitation pulse representing each pulse excitation vector together with the pulse excitation vectors 1 and 2 to the selector 3005. Is output. Further, the selector 3005 outputs information indicating which one of the pulse excitation vectors 1 and 2 has been selected, and the polarity and index of each pulse representing the selected pulse excitation vector, to the outside of the excitation generator. The selection information and the polarity and index information of the excitation pulse are converted into a data sequence to be output to the transmission path through an encoder, a multiplexer and the like, and sent out to the transmission path.

Adder 3009 performs vector addition of the adaptive code vector component output from multiplier 3007 and the pulse excitation vector component output from multiplier 3008, and outputs the result as an excitation excitation vector.

In this embodiment, as in Embodiment 12 or Embodiment 13 or Embodiment 14, index updating means or pulse number and index updating means or a combination of fixed search position and phase adaptive search position is used. Is provided before the pulse position searcher 3004, it is possible to reduce the property of being easily affected by a transmission path error caused by using the search position calculator 3003.

As for the method of setting pulses, in the case where a constant number of pulses, for example, four pulses are set in a search range, for example, somewhere in 32 positions, 32 positions are divided into four as described above, and one pulse is set. All combinations (8 × 8) are determined so that one pulse is determined to be one of eight assigned pulse positions.
In addition to the method of searching (× 8 × 8 ways), there is a method of searching for all the combinations that select four places from 32 places. In addition to the combination of the impulses having the amplitude of 1, a combination of a plurality of pulses, for example, two pulses, or a combination of impulses having different amplitudes may be used.

In a mode in which the number of pulses is small and the pulse position information is sufficient, a portion of the pulse position information is assigned to an index representing a noise code vector within a range where the pulse position information is not insufficient, so that voiced start-up is performed. It is possible to improve the performance not only for the unvoiced part but also for the unvoiced consonant part and the noise-like input signal.

Further, the functions of the audio encoding apparatus described in the first to seventeenth embodiments can be recorded as a program on a recording medium such as a magnetic disk, a magneto-optical disk, an IC card, a ROM, and a RAM. Therefore, by reading the recording medium with a computer, the function of the audio encoding device can be realized.

[0225]

According to the present invention, as is apparent from the above embodiment, an amplitude emphasis window for emphasizing the amplitude of a noise code vector corresponding to a pitch peak position of an adaptive code vector is multiplied by the noise code vector. Therefore, the sound quality can be improved by using the phase information existing in one pitch waveform.

In the present invention, since the noise code vector limited only to the vicinity of the pitch peak of the adaptive code vector is used, even if the number of bits allocated to the noise code vector is small, the sound quality degradation can be reduced and the pitch peak can be reduced. It is possible to improve the voice quality of a voiced part where power is concentrated in the vicinity.

According to the present invention, the search range of the pulse position is determined based on the pitch peak position and the pitch cycle of the adaptive code vector. Therefore, the pulse position search according to the pitch cycle within one pitch waveform is performed. Therefore, even when the number of bits allocated to the pulse position is small, it is possible to suppress the deterioration of the voice quality.

According to the present invention, a sound source signal having a pitch periodicity can be efficiently represented by limiting the range of the pulse search to a length slightly longer than one pitch period. Also, since two pitch peaks are included in the search range, it is possible to handle cases where the shape of the first pitch peak is different from the shape of the second pitch peak, or where the position of the first pitch peak is erroneously detected. It is.

The present invention also has a configuration in which the number of pulses is adaptively changed in accordance with the pitch period of the input audio signal. Can be planned.

According to the present invention, since the pulse amplitude in the vicinity of the pitch peak and in the other portion is determined before the pulse position search, the shape of one pitch waveform can be efficiently expressed.

According to the present invention, a pulse sound source search suitable for a voiced rising part / unvoiced part and a voiced stationary part / voiced part is performed by switching the pulse search position using the continuity of the pitch period. Therefore, the sound quality can be improved.

In the present invention, the first stage quantization is performed on the pitch gain (adaptive code vector gain) of the current subframe by using the pitch gain obtained immediately after the adaptive codebook search, and the optimum gain obtained at the end of the excitation search is obtained. In a CELP-type speech coding apparatus that generates a driving excitation vector by the sum of an adaptive codebook and a fixed codebook (noise codebook) by quantizing the difference between the pitch gain and the first-stage quantization pitch gain in the second step, Since the information obtained before the fixed codebook (noise codebook) search is quantized and transmitted, it is possible to switch the fixed codebook (noise codebook) without adding independent mode information, Audio information can be efficiently encoded.

The present invention also determines the pitch periodicity of the audio signal of the current subframe based on the continuity of the previously encoded pitch period or the magnitude (or continuity) of the previously encoded pitch gain. However, since the search position of the pulse sound source is switched, the pulse sound source search suitable for each part can be performed without adding new information to the determination of the high and low pitch periodicity. Thus, it is possible to improve the voice quality under the same information amount.

According to the present invention, the pitch peak position in the current subframe is predicted backward by using the pitch peak position in the immediately preceding subframe, the pitch period in the immediately preceding subframe, and the pitch period in the current subframe. Since it is possible to switch whether or not to perform the phase adaptation process using the predicted pitch peak position, it is possible to switch the phase adaptation process without newly transmitting the switching information, and to improve the voice quality under the same information amount. Improvement can be achieved. In a mode in which the phase adaptation process is not performed, a fixed codebook may be used.
The occurrence of a state where the fixed codebook is continuously used in a silent part or the like occurs, whereby the effect of resetting the propagation of an error to the phase adaptive excitation can be obtained.

In the present invention, since the phase adaptation is switched by using the signal power concentration in the vicinity of the pitch peak of the adaptive code vector, the phase adaptation process is switched without newly transmitting switching information. Therefore, it is possible to improve the voice quality under the same information amount. In the mode in which the phase adaptation process is not performed, the fixed codebook may be used, and a state in which the fixed codebook is continuously used in a silent part or the like occurs. Can be obtained.

The present invention also provides a CELP-type speech coding apparatus that expresses the position of an excitation pulse by a relative position where the pitch peak position is 0, so that indices indicating the positions of the excitation pulse are arranged in order from the head of the subframe. In the case where the pitch peak position is erroneously set due to the influence of a transmission path error or the like, the deviation of the sound source pulse position can be prevented from becoming very large.

According to the present invention, in a CELP type speech coding apparatus which expresses the position of an excitation pulse by a relative position where a pitch peak position is 0, an index indicating each position of an excitation pulse is arranged in order from the head of a subframe. In addition to the above, when the pitch peak position is erroneously set due to the influence of a transmission line error, etc., by defining the numbers to be attached to the different pulses represented by the same index number in order from the head of the subframe. , The displacement of the sound source pulse position can be reduced.

The present invention also provides a CELP-type speech coding apparatus which expresses the position of an excitation pulse by a relative position where the pitch peak position is 0, instead of expressing all the search positions of the excitation pulse by a relative position. By expressing only a part as a relative position and setting the remaining search position to a predetermined fixed position, when the pitch peak position is incorrect due to the influence of a transmission path error, the position of the sound source pulse is shifted. By reducing the probability of occurrence, it is possible to prevent the effect of the transmission path error from propagating for a long time.

According to the present invention, since a peak position within one pitch waveform is searched for as a pitch peak position, it is possible to prevent erroneous detection in which a second peak caused by a mismatch between a subframe length and a pitch period is regarded as a pitch peak. it can.

According to the present invention, in a continuous voiced stationary part, the current pitch peak is obtained by using information on the position of the pitch peak in the immediately preceding subframe, the pitch period in the immediately preceding subframe, and the pitch period in the current subframe. By limiting the existing range of the position and searching for the pitch peak position within the range, the second peak in one pitch waveform generated when searching for the pitch peak position using only the signal of the current subframe is obtained. It is possible to prevent erroneous detection using a peak as a pitch peak.

The present invention also provides a CELP-type speech coding apparatus in which a pulse excitation is applied to a noise codebook, in a mode in which the number of excitation pulses is small and the position information of each excitation pulse is sufficient. The noise codebook configuration has both a mode with a large number of excitation pulses instead of a coarse, so that it is possible to achieve both an improvement in the voice quality of voiced rising portions and an effective use of the mode with a large number of excitation pulses. .

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a sound source generation unit of a CELP speech coding apparatus according to a first embodiment of the present invention.

FIG. 2 is a schematic diagram illustrating the relationship between the shape of an amplitude enhancement window, an adaptive code vector, and a pitch pulse position according to the first embodiment of the present invention.

FIG. 3 shows C in a modification of the first embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the ELP speech encoding device.

FIG. 4 is a block diagram showing a configuration of a sound source generation unit of a CELP speech coding apparatus according to a second embodiment of the present invention.

FIG. 5 is a block diagram illustrating a configuration of a sound source generation unit of a CELP speech coding apparatus according to a third embodiment of the present invention.

FIG. 6 is a schematic diagram showing an arrangement of a pulse position vicinity limited vector according to a third embodiment of the present invention.

FIG. 7 is a schematic diagram showing an arrangement of a pulse position neighborhood limited vector according to the third embodiment of the present invention (continued).

FIG. 8 is a block diagram showing a configuration of a sound source generation unit of a CELP speech coding apparatus according to a fourth embodiment of the present invention.

FIG. 9 is a schematic diagram showing a pulse sound source search range according to a fourth embodiment of the present invention.

FIG. 10 is a schematic diagram showing a pulse sound source search range according to the fourth embodiment of the present invention (continued).

11A is a block diagram illustrating a configuration of a search position calculator according to a fifth embodiment of the present invention. FIG. 11B is a schematic diagram illustrating an example of a pulse search position pattern.

FIG. 12 shows CELP according to a sixth embodiment of the present invention.
Block diagram showing a configuration of a sound source generation unit of the type speech coding apparatus

13A to 13D are schematic diagrams illustrating an example of a pulse search position calculated by a search position calculator according to the sixth embodiment of the present invention.

FIG. 14 shows a CELP according to a seventh embodiment of the present invention.
Block diagram showing a configuration of a sound source generation unit of the type speech coding apparatus

FIG. 15 shows a CELP according to an eighth embodiment of the present invention.
Block diagram showing a configuration of a sound source generation unit of the type speech coding apparatus

FIGS. 16A and 16B are list views showing examples of fixed search position patterns used in the eighth embodiment of the present invention.

FIG. 17 shows a CELP according to a ninth embodiment of the present invention.
Block diagram showing a configuration of a sound source generation unit of the type speech coding apparatus

FIG. 18 shows a CEL according to a tenth embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the P-type speech encoding device.

FIG. 19 is a schematic diagram illustrating a prediction principle in a pitch peak position predictor according to a tenth embodiment of the present invention.

FIG. 20 shows a CEL according to an eleventh embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the P-type speech encoding device.

FIG. 21 shows a CEL according to a twelfth embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the P-type speech encoding device.

FIG. 22 shows a search position pattern of a sound source pulse output by a search position calculator in the twelfth embodiment of the present invention, an index corresponding to each position when no index updating means is provided, and an index updating means. Schematic diagram showing the index corresponding to each position when equipped

FIG. 23 shows a CEL according to a thirteenth embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the P-type speech encoding device.

FIG. 24 (a) is a schematic diagram showing a pattern of a sound source pulse search position outputted by a search position calculator in a thirteenth embodiment of the present invention, and a correspondence between a relative position and an absolute position corresponding to each position. Schematic diagram showing a pulse number and an index assigned to each sound source pulse in a case where the means for updating a pulse number and an index in the thirteenth embodiment of the present invention is not provided. (C) In the thirteenth embodiment of the present invention Schematic diagram showing pulse numbers and indices assigned to each sound source pulse when a pulse number and index updating means are provided.

FIG. 25 shows a CEL according to a fourteenth embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the P-type speech encoding device.

FIGS. 26A and 26B are schematic diagrams showing an example of a fixed search position pattern used in the fourteenth embodiment of the present invention. FIGS. 26B and 26C are search positions used in the fourteenth embodiment of the present invention. Schematic diagram illustrating an example of a pattern of a sound source pulse search position generated by a calculator. (D) Schematic diagram illustrating an example of a pattern of a sound source pulse search position used in the pulse position search device according to the fourteenth embodiment of the present invention.

FIG. 27 shows a CEL according to a fifteenth embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the P-type speech encoding device.

FIGS. 28A and 28B are schematic diagrams showing an example of an adaptive code vector waveform in which a pitch peak and a second peak are incorrect in a pitch peak calculator. FIG. Schematic diagram showing an example of the illustrated adaptive code vector waveform

FIG. 29 shows a CEL according to a sixteenth embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the P-type speech encoding device.

FIG. 30 shows a CEL according to a seventeenth embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the P-type speech encoding device.

FIG. 31 is a block diagram showing a configuration of a sound source generation unit of a conventional general CELP speech coding apparatus.

FIG. 32 shows a conventional noise source having a pitch periodicization unit C
FIG. 2 is a block diagram illustrating a configuration of a sound source generation unit of the ELP speech encoding device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 11 Adaptive codebook 12 Pitch peak position calculator 13 Amplitude emphasis window generator 14 Noise codebook 15 Periodizer 16 Amplitude emphasis window multiplier 21 Adaptive codebook 22 Pitch peak position calculator 23 Amplitude emphasis window generator 24 Noise codebook 25 Amplitude emphasis window multiplier 31 Pulse train sound source 32 Amplitude emphasis window generator 33 Adder 34 Noise source 35 Multiplier 41 Adaptive codebook 42 Phase searcher 43 Pitch pulse position neighborhood limited noise codebook 44 Noise vector generator 45 Periodization Unit 51 adaptive codebook 52 pitch peak position calculator 53 search range calculator 54 pulse position searcher 55 pitch gain multiplier 56 pulse excitation gain multiplier 57 adder 61 pulse search position pattern selector 62 pulse search position determiner 71 adaptation Codebook 72 Pitch peak position calculator 73 Pulse number determiner 7 Search position calculator 75 Pulse position searcher 76 Multiplier 77 Multiplier 78 Adder 81 Adaptive codebook 82 Pitch peak position calculator 83 Number of pulses determiner 84 Search position calculator 85 Pulse position searcher 86 Adder 87 Pulse amplitude calculation Unit 88 multiplier 89 multiplier 90 adder 91 adaptive codebook 92 pitch peak position calculator 93 pulse number determiner 94 search position calculator 95 delay unit 96 determiner 97 pulse position searcher 98 switch 99 multiplier 100 multiplier 101 Adder 111 adaptive codebook 112 pitch peak position calculator 113 pulse number determiner 114 search position calculator 115 switch 116 pitch gain calculator 117 quantizer 118 determiner 119 pulse position searcher 120 adder 121 difference quantizer 122 Adder 123 Multiplier 124 Arithmetic unit 125 adder 1801 adaptive codebook 1802 pitch peak position calculator 1803 delay unit 1804 delay unit 1805 pitch peak position predictor 1806 determiner 1807 search position calculator 1808 switch 1809 pulse position searcher 1810 multiplier 1811 adder 1812 multiplication 2001 Adaptive codebook 2002 Pitch peak position calculator 2003 Pulsability determinator 2004 Search position calculator 2005 Switch 2006 Pulse position searcher 2007 Multiplier 2008 Adder 2009 Multiplier 2101 Adaptive codebook 2102 Pitch peak position calculator 2103 Search position Calculator 2104 Index updating means 2105 Pulse position searcher 2106 Multiplier 2107 Multiplier 2108 Adder 2301 Adaptive codebook 2302 Pitch peak position calculation 2303 Search position calculator 2304 Pulse number and index updating means 2305 Pulse position searcher 2306 Multiplier 2307 Multiplier 2308 Adder 2501 Adaptive codebook 2502 Pitch peak position calculator 2503 Search position calculator 2504 Adder 2505 Pulse position search 2506 Multiplier 2507 Multiplier 2508 Adder 2701 Adaptive codebook 2702 Pitch peak position calculator 2703 Pitch peak position corrector 2704 Search position calculator 2705 Pulse position searcher 2706 Multiplier 2707 Multiplier 2708 Adder 2901 Adaptive codebook 2902 Pitch peak position calculator 2903 Pitch peak search range limiter 2904 Delay device 2905 Delay device 2906 Search position calculator 2907 Pulse position search device 2908 Multiplier 2 09 multipliers 2910 adder 3001 adaptive codebook 3002 pitch peak position calculator 3003 search position calculator 3004 pulse position searcher 3005 selector 3006 pulse position searcher 3007 multiplier 3008 multiplier 3009 adder

Claims (40)

[Claims] 1. A speech encoding apparatus comprising: a CELP-type speech encoding apparatus; and a sound source generation unit for enhancing an amplitude of a noise code vector corresponding to a pitch peak position of an adaptive code vector. 2. A sound source generating unit which multiplies a noise code vector by an amplitude emphasis window synchronized with a pitch period of the adaptive code vector, thereby enhancing the amplitude of the noise code vector corresponding to the position of the pitch peak of the adaptive code vector. The speech encoding device according to claim 1. 3. The speech encoding apparatus according to claim 2, wherein the sound source generation unit uses a triangular window centered on a pitch peak position of the adaptive code vector as an amplitude emphasizing window. 4. A speech encoding apparatus comprising a CELP-type speech encoding apparatus and an excitation generation unit using a noise code vector limited only to a vicinity of a pitch peak of an adaptive code vector. 5. A CEL using a pulse excitation as a noise codebook
A P-type speech encoding device, comprising a sound source generation unit that determines a search range of a pulse position by a pitch period and a pitch peak position of an adaptive code vector. 6. The speech coding apparatus according to claim 5, wherein the sound source generation unit determines the search range of the pulse position so that the vicinity of the pitch peak position of the adaptive code vector is dense and the other parts are sparse. 7. The speech coding apparatus according to claim 5, wherein a search range of a pulse position is switched according to a pitch period. 8. The speech coding according to claim 7, wherein when a plurality of pitch peaks are present in the adaptive code vector, the search range of the pulse position is limited so that at least two pitch peak positions are included in the search range. apparatus. 9. A CLP-type speech coding apparatus having a configuration in which a noise codebook is switched according to a speech analysis result.
ELP type speech coding device. 10. A CELP-type speech coding apparatus,
A speech coding apparatus including a sound source generation unit that switches a noise codebook using transmission parameters extracted before performing a noise codebook search. 11. The speech encoding apparatus according to claim 5, further comprising a sound source generation unit that switches the number of pulses according to a result of the speech signal analysis. 12. The speech according to claim 5, further comprising a sound source generation unit that switches the number of pulses using a transmission parameter extracted before performing a random codebook search. Encoding device. 13. The speech encoding apparatus according to claim 5, further comprising a sound source generation unit that switches the number of pulses according to a pitch period. 14. The speech coding apparatus according to claim 13, wherein the number of pulses is switched between a case where the fluctuation of the pitch period between successive subframes is small and a case where the fluctuation is not so. 15. The noise code vector generation unit using a pulsed sound source as a noise source determines the pulse amplitude prior to the pulse position search. A speech encoding device according to claim 1. 16. The speech coding apparatus according to claim 15, wherein the noise code vector generation unit using a pulse sound source as the noise sound source changes the pulse amplitude in the vicinity of the pitch peak of the adaptive code vector and in other parts. 17. The speech encoding apparatus according to claim 13, wherein the number of pulses of a pulse sound source to be used is determined based on a pitch period, either statistically or by learning. 18. A CELP-type speech coding apparatus, comprising:
Equipped with an excitation generator that quantizes the pitch gain in multiple stages, the first stage uses the value obtained immediately after the adaptive codebook search as the quantization target, and the second and subsequent stages are determined by closed loop search after completing all the excitation search An audio encoding device that uses a difference between a pitch gain and a value quantized in a first stage as a quantization target. 19. The fixed codebook is switched using the quantization value of the pitch gain obtained immediately after the adaptive codebook search of the speech coding apparatus according to claim 18.
The speech encoding device according to claim 2 or any one of claims 15 to 17. 20. The fixed codebook is switched based on a change in pitch period between subframes.
20. The speech encoding device according to any one of claims 15 to 19. 21. The speech encoding apparatus according to claim 9, wherein the fixed codebook is switched using the pitch gain quantized in the immediately preceding subframe. 22. The speech coding apparatus according to claim 9, wherein the fixed codebook is switched based on a change in pitch period between sub-frames and a quantization pitch gain. . 23. The speech coding apparatus according to claim 19, wherein a pulse excitation codebook is used as the fixed codebook. 24. A CELP-type speech coding apparatus that performs a speech coding process for each subframe having a predetermined time length, and determines whether a phase in a current subframe and a phase in a previous subframe are continuous. An audio encoding device that determines and switches a sound source to be used when it is determined to be continuous and when it is determined to be not continuous. 25. Predict a pitch peak position in a current subframe using a pitch peak position in the immediately preceding subframe, a pitch period in the immediately preceding subframe, and a pitch period in the current subframe.
Depending on whether the pitch peak position in the current subframe obtained by this prediction is close to the pitch peak position obtained only from data in the current subframe, the phase in the immediately preceding subframe and the phase in the current subframe are 26. The CELP-type speech encoding apparatus according to claim 24, wherein it is determined whether or not the sound source is continuous, and the encoding processing method of the excitation is switched according to the determination result. 26. When it is determined that the phase in the immediately preceding subframe and the phase in the current subframe are continuous, phase adaptive processing is performed on the noise codebook, and the phase in the immediately preceding subframe is determined. 26. The speech encoding apparatus according to claim 24, wherein when it is determined that the phase in the current subframe is not continuous with the phase in the current subframe, the phase adaptation processing is not performed on the noise codebook. 27. A CELP-type speech coding apparatus for performing speech coding processing for each subframe having a predetermined time length, based on a signal power concentration near a pitch peak of an adaptive code vector in a current subframe. , An audio encoding device that switches an encoding method of an excitation signal. 28. When the ratio of the signal power in the vicinity of the pitch peak of the adaptive code vector in the current subframe to the entire signal of one pitch cycle length is equal to or greater than a predetermined value, the phase adaptation processing is performed in the noise codebook. Done against
28. The speech encoding apparatus according to claim 27, wherein when the value is less than a predetermined value, the phase adaptation processing is not performed on the noise codebook. 29. A pulse sound source is applied to a noise sound source as a phase adaptation process, wherein a pulse position search is performed densely near a pitch peak and a pulse position search is performed sparsely in a portion other than the vicinity of a pitch peak. Item 30. The audio encoding device according to Item 28. 30. An index representing the position of a pulse is
30. The speech encoding apparatus according to claim 5, wherein the speech encoding apparatus is arranged so as to be arranged in order from the head side of the subframe. 31. In the case of the same index number, pulse numbers are assigned in order from the head of the subframe, and each pulse is so dense that the vicinity of the pitch peak position is dense and the parts other than the vicinity of the pitch peak are sparse. 31. The speech encoding device according to claim 30, wherein a search position is determined. 32. A part of the pulse search position is determined by the pitch peak position, and the other pulse search positions are fixed positions that are predetermined regardless of the pitch peak position.
Claims 5 to 8 or Claims 11 to 17
30. The speech encoding device according to claim 23. 33. A pitch peak position calculation for determining a pitch peak position of a voice or sound source signal having a predetermined time length, extracting only one pitch period length from the signal and determining a pitch peak position in the extracted signal. The speech encoding apparatus according to any one of claims 1 to 8, or 11 to 17, or 19 to 23, or 25 to 32, comprising means. 34. When extracting only one pitch period length from the signal, first determine a pitch peak position using the entire signal without extracting one pitch period length, and start extracting the determined pitch peak position. 34. The speech encoding apparatus according to claim 33, wherein one pitch period length is cut out as a point, and a pitch peak position is determined in the cut out signal. 35. A CELP-type speech coding apparatus that performs a speech coding process for each subframe having a predetermined time length, calculates a pitch peak position in a current subframe, and calculates a pitch period in a preceding subframe. If the difference from the pitch cycle in the current subframe is within a predetermined range, the pitch peak position in the immediately preceding subframe, the pitch cycle in the immediately preceding subframe, and the pitch cycle in the current subframe are used. Predict the pitch peak position in the current sub-frame, and using the pitch peak position in the current sub-frame obtained by this prediction to limit the range of the pitch peak position in the current sub-frame in advance,
Speech coding according to any one of claims 1 to 8, claim 11 to claim 17, or claim 19 to claim 23, or claim 25 to 32, wherein a pitch peak position search is performed within the range. apparatus. 36. A CELP-type speech coding apparatus for performing speech coding processing for each subframe having a predetermined time length, wherein a pulse excitation is used as a noise codebook, and the noise codebook has at least two or more modes. The number of sound source pulses can be changed by switching the mode. At least one is a mode in which the position information of each pulse is sufficient and the number of pulses is small, and the other is a mode in which the position information of each pulse is insufficient. A speech encoding device that has many modes and switches modes by transmitting mode switching information. 37. When the pitch period is short, the search range of the source pulse is limited to a narrow range corresponding to the pitch period, thereby reducing the position information of the source pulse and increasing the number of source pulses. 36. The speech encoding device according to 36. 38. In a mode in which the position information of each pulse is insufficient but the number of pulses is large, the search position of the sound source pulse is made dense near the pitch peak position, and the search position of the sound source pulse is made sparse in other portions. 38. The speech encoding device according to claim 36, wherein the search range of the pulse position is determined so as to be as follows. 39. A CELP-type speech coding apparatus according to claim 36, wherein in a sound source mode in which the number of pulses is small and the position information is sufficient, a part of the position information is converted to a noise source. An audio encoding device that is assigned to an index representing a code vector. 40. A computer-readable storage medium storing a program for executing the functions of the speech encoding device according to claim 1. Description: