JPH06202699A - Speech encoding device and speech decoding device, and speech encoding and decoding method - Google Patents

Abstract

PURPOSE:To prevent sound quality deterioration due to the disorder of the periodicity of a synthesized speech by a small increase in the amount of transmitted information as to the speech encoding and decoding devices which separate an input speech signal into spectrum envelope information and sound source signal information and quantize the sound source signal into vectors. CONSTITUTION:A periodic pulse sound source generating means 21 which generates a sound source vector consisting of a periodic pulse train corresponding uniformly to an adaptive sound source vector outputted from an adaptive sound source code book consisting of a sound source signal found in a precedent frame and a multiple vector adaptive sound source encoding means 20 which generates a sound source signal by adding gains individually to adaptive sound source vectors and periodic pulse sound source vectors corresponding to the adaptive sound source vectors, and finds and encodes pitch information and gains so as to obtain a sound source signal for generating a synthesized speech minimizing distortion from the input speech are provided on an encoding side, and the same periodic pulse sound source generating means 23 as the encoding side and a multiple vector adaptive sound source decoding means 22 which generates adaptive sound source vectors and periodic pulse sound source vectors from the transmitted codes are provided on a decoding side.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice encoding / decoding device which is used when digitally transmitting or storing voice, and which vector-quantizes voice source signal information.

[0002]

2. Description of the Related Art A conventional speech coding / decoding apparatus for separating speech into spectral envelope information and excitation signal information and vector-quantizing the excitation signal information is shown in FIG. Figure 10 shows WBKleijn, DJKrasinski, and RHK
"Improved Speech Quality and Efficient V" by etchum
ector Quantization in SELP ”(ICASSP ■ 88, pp.155-1
58, 1988). In the figure, 1 is an encoding unit, 2 is a decoding unit, 3 is a transmission path,
Reference numeral 4 is an input voice, and 5 is an output voice. 6 is spectrum analysis means, 7 is spectrum parameter coding means,
Reference numeral 8 is an adaptive excitation coding means, and 9 and 14 are adaptive excitation codebooks. 10 is a driving excitation coding means, and 11 and 16 are driving excitation codebooks. 12 and 17 are sound source vector generation means,
Reference numeral 13 is an adaptive excitation decoding means, 15 is a driving excitation decoding means, 18 is a spectrum parameter decoding means, 19
Is a synthesis filter.

The operation of the conventional speech encoding / decoding device will be described below. First, the operation of the encoding unit 1 will be described. The spectrum parameter analysis means 6 analyzes the input voice 4 and extracts spectrum parameters. The spectrum parameter coding means 7 quantizes the spectrum parameter, outputs the code corresponding to the quantized spectrum parameter to the decoding section 2 via the transmission path 3, and also outputs the quantized spectrum parameter to the adaptive excitation coding means 8 and the driving excitation code. It outputs to the conversion means 10. The excitation signal obtained in the preceding frame is stored in the adaptive excitation codebook 9, and the adaptive excitation vector obtained by repeating the excitation signal in the pitch cycle in the current frame is output using the pitch information as a code. FIG. 11 shows an example of the sound source signal in the preceding frame and the adaptive sound source vector obtained at that time. When the pitch period is shorter than the frame length, the adaptive excitation vector is a repetition of the excitation signal in the current frame at the pitch period as shown in FIG. 11 (b).

The adaptive excitation coding means 8 has M adaptive excitation vectors a _i (i) input from the adaptive excitation codebook 9.
= 1, ..., M) and the spectrum parameter input from the spectrum parameter encoding means 7,
M synthesized speech vectors A _i (i = 1, ..., M) are synthesized. Then, the inter-vector distance D _i between the input voice 4 and the input voice vector X cut out for each frame is obtained, for example, according to the equation (1). Here, the gain β _i is determined, for example, according to the equation (2) so that the distance D _i becomes the minimum. (Note that t indicates transposition)

[0005]

[Equation 1]

Next, a vector A _I that minimizes the inter-vector distance is searched for, and the code of the gain _{I qI obtained} by quantizing the code I and the gain β _I is output to the decoding unit 2 via the transmission line 3. In addition, the selected adaptive _excitation vector a _I and its gain β _qI are output to the _excitation vector generation means 12, and the error vector X ■ X ■ = X−β _qI A _I is output to the driving _excitation encoding means 10. To do.

The driving excitation codebook 11 stores N driving excitation vectors generated from random noise, for example. The driving excitation coding means 10 is a driving excitation vector c _j (j = 1, ...) Inputted from the driving excitation codebook 11.
, N) and the spectrum parameter input from the spectrum parameter coding means, N synthesized speech vectors C _j (j = 1, ..., N) are synthesized. Then, the inter-vector distance D _j with the error vector X ■ input from the adaptive excitation encoding means 8 is calculated by, for example, the equation (3)
Ask according to. Here, the gain γ _j is determined, for example, according to the equation (4) so that the distance D _j is minimized.

[0008]

[Equation 2]

Next, a vector C _J having the minimum inter-vector distance is searched for, and the code J and the gain γ _{qJ obtained} by quantizing the gain γ _J are output to the decoding unit 2 via the transmission line 3. At the same time, the selected driving sound source vector c _J and its gain γ _qJ are output to the sound source vector generating means 12.

The _excitation vector generation means 12 has the adaptive _excitation vector _aI and its gain β _qI input from the adaptive _excitation coding means 8 and the driving _excitation vector c _J and its input from the driving _excitation coding means 10. An _excitation vector β _qI a _I + γ _qJ c _J is generated from the gain γ _qJ , and this is output to the adaptive excitation codebook 9.

Next, the operation of the decoding section 2 will be described. Based on the code I of the adaptive excitation vector input from the encoding unit 1, the adaptive excitation decoding means 13 stores the same adaptive excitation vector as the adaptive excitation codebook 9 from the adaptive excitation codebook 14 that stores the adaptive excitation vector a. _I is read, the gain β _qI is decoded from the sign of the gain for the adaptive _excitation vector input from the encoding unit 1, and the adaptive _excitation vector a _I and its gain β _qI are output to the _excitation vector generation means 17. . Further, the driving sound source decoding means 15
Based on the code J of the drive excitation vector input from the encoding unit 1, the drive excitation vector cJ is read from the drive excitation codebook 16 that stores the same drive excitation vector as the drive excitation codebook 11, and is also encoded. The gain γ _qJ is decoded from the sign of the gain for the driving sound source vector input from the unit 1, and the driving sound source vector c _J and its gain γ _qJ are output to the sound source vector generating means 17.

The excitation vector generation means 17 is an adaptive excitation vector a _I input from the adaptive excitation decoding means 13.
And its gain β _qI , the driving _excitation vector c _J input from the driving _excitation decoding means 15, and its gain γ.
_{An excitation} vector β _qI a _I + γ _qJ c _J is generated from _qJ and is output to the adaptive _excitation codebook 14 and the synthesis filter 19.

The spectrum parameter decoding means 18 is
The spectrum parameter is decoded based on the code of the spectrum parameter input from the encoding unit 1 and output to the synthesis filter 19. The synthesis filter 19 synthesizes the output speech 5 using the excitation vector input from the excitation vector generation means 17 and the spectrum parameter input from the spectrum parameter decoding means 18.

[0014]

Voice signals include voiced sound and unvoiced sound, and are characterized by whether the sound source signal has a pulse-like component of a pitch period or white noise having no periodicity. The reproducibility of pitch periodicity in voiced sound has a great influence on the quality of synthesized speech. In the conventional speech encoding / decoding apparatus as described above, the pulse-like component of the pitch period of the excitation signal in the voiced sound is generated by using the adaptive excitation vector created by the excitation signal of the preceding frame. However, since the adaptive sound source vector is composed of multiple samples, it is not possible to positively generate the pulse-like component necessary for voiced sound. Therefore, there is a problem that an appropriate sound source cannot be generated in the transition part from unvoiced sound to voiced sound and the sound quality is deteriorated. In addition, even if a pulse component is generated, the sound source is generated by including the other samples, so that an optimum pulse sequence cannot always be obtained, which leads to lack of pitch periodicity, resulting in deterioration of sound quality. It was

Further, since the sound source signal is generated by a linear sum of the adaptive sound source vector and the driving sound source vector in which the pitch periodicity is not taken into consideration, even if the adaptive sound source vector generates the pitch periodicity, the transmitted information amount is In order to reduce it, it is necessary to increase the frame length, that is, the vector length, or reduce the codebook size of the driving excitation codebook. In that case, the pitch periodicity of the generated excitation signal is disturbed and the sound quality is reduced. There was also the problem of deterioration.

In the conventional speech coding / decoding apparatus, the gain quantization for the driving excitation vector is performed regardless of the power of the input speech, but since the range of gain to be quantized is large, the quantization is performed with a small number of bits. If so, the quantization distortion becomes large and the quality of the synthesized speech deteriorates. Further, in order to change the quantization of the gain of the driving sound source vector by the power of the input sound, it is necessary to newly transmit the power information, which is not efficient.

The present invention has been made to solve the above problem, and in addition to an adaptive excitation vector, generates a excitation vector consisting of a periodic pulse train that can be uniquely obtained for this adaptive excitation vector, By using this to generate a sound source signal and by repeatedly using the driving sound source vector in synchronization with the pitch cycle, the pitch periodicity of the sound source signal is not disturbed even with a small amount of transmission information, and the power of the adaptive sound source signal is reduced. By changing the quantization of the gain of the driving excitation vector according to the information, it is an object to synthesize high-quality decoded speech with small quantization distortion of the excitation signal even with a small amount of transmission information.

[0018]

A speech coding apparatus according to the present invention includes an adaptive excitation codebook for storing an adaptive excitation vector which is information of an excitation signal of a preceding frame, and each adaptive excitation in the adaptive excitation codebook. Periodic pulse source generating means for generating a periodic pulse source vector determined as a main component of the vector, and a decoded speech generated as a source signal by a linear sum of the adaptive source vector and the periodic pulse source vector corresponding to the adaptive source vector and the input An adaptive excitation vector that minimizes distortion with speech and a periodic pulse excitation vector corresponding to the adaptive excitation vector, and a gain for the adaptive excitation vector and the periodic pulse excitation vector are obtained,
A plurality of vector adaptive excitation encoding means for encoding the adaptive excitation vector and the gains for the adaptive excitation vector and the periodic pulse excitation vector are provided.

The decoding apparatus according to the second aspect of the present invention is an adaptive excitation codebook for storing an adaptive excitation vector which is information of an excitation signal of a preceding frame, and a periodic pulse determined as a main component of the decoded adaptive excitation vector. A periodic pulse sound source generating means for generating a sound source vector, an adaptive sound source vector, the adaptive sound source vector, and a gain for the periodic pulse sound source vector corresponding to the adaptive sound source vector are decoded to obtain the adaptive sound source vector and the periodic pulse sound source vector. A multi-vector adaptive excitation decoding means for adding the above gains to each of them and adding them.

An encoding apparatus according to a third aspect of the present invention inputs an adaptive excitation codebook that stores an adaptive excitation vector that is information of an excitation signal of a preceding frame, and a decoded speech generated by using the adaptive excitation vector as an excitation signal. An adaptive sound source vector that minimizes distortion with speech and an adaptive sound source coding unit that obtains a gain for the adaptive sound source vector and encodes the adaptive sound source vector and the gain for the adaptive sound source vector, and a plurality of previously prepared sound sources A drive excitation codebook that stores a signal as a drive excitation vector, and a plurality of feature points that are arranged at pitch cycle intervals with respect to the input voice as pitch positions, and a pitch position extraction unit that encodes the pitch positions, Align the predetermined position of each driving excitation vector of the driving excitation codebook to the pitch position, and pitch the driving excitation vector. And a pitch synchronization means for generating a pitch-synchronized drive source vector repeated by the above, and a distortion between a decoded voice generated as a source signal and a linear sum of the adaptive excitation vector and the pitch-synchronized drive source vector and an input voice are minimized. A pitch-synchronized drive excitation vector and a gain for the pitch-synchronized drive excitation vector are obtained, and a drive excitation vector corresponding to the pitch-synchronized drive excitation vector and a drive excitation encoding means for encoding the gain for the pitch-synchronized drive excitation vector are provided. It was

An encoding apparatus according to a fourth aspect of the present invention includes an adaptive excitation codebook that stores an adaptive excitation vector that is information of an excitation signal of a preceding frame, and a decoded speech generated by generating an excitation signal of the adaptive excitation vector. An adaptive excitation vector that minimizes distortion with respect to the input speech, an adaptive excitation encoding unit that obtains a gain for the adaptive excitation vector, and encodes the adaptive excitation vector and the gain for the adaptive excitation vector, and a plurality of previously prepared adaptive excitation encoding units. A driving excitation codebook that stores excitation signals as driving excitation vectors and a plurality of points arranged at pitch period intervals as pitch positions, and the predetermined positions of the driving excitation vectors of the driving excitation codebook are aligned with the pitch positions. A pitch synchronization means for repeating the driving sound source vector in a pitch cycle to generate a pitch synchronization driving sound source vector; A pitch-synchronized drive source vector that minimizes distortion between the decoded voice and the input voice generated by using a linear sum of the excitation vector and the pitch-synchronized drive source vector as a source signal and a gain for the pitch-synchronized drive source vector are obtained, and the pitch A drive excitation vector corresponding to the synchronous drive excitation vector, a pitch position, and a drive excitation encoding means for encoding a gain for the pitch synchronous drive excitation vector are provided.

A decoding device according to a fifth aspect of the present invention decodes an adaptive excitation codebook that stores an adaptive excitation vector that is information of an excitation signal of a preceding frame, an adaptive excitation vector and a gain for the adaptive excitation vector. Adaptive excitation decoding means, a driving excitation codebook that stores a plurality of excitation signals prepared in advance as driving excitation vectors, a pitch position is decoded, and a predetermined driving excitation vector of the decoded driving excitation vector is obtained at the pitch position. Pitch synchronization means for aligning positions and generating the pitch-synchronized driving sound source vector by repeating the driving sound source vector in a pitch cycle, and the driving sound source vector,
A sound source from a driving sound source decoding unit that decodes a gain for the pitch synchronization driving sound source vector, a gain for the adaptive sound source vector and the adaptive sound source vector, and a gain for the pitch synchronization driving sound source vector and the pitch synchronization driving sound source vector Excitation vector generation means for decoding the vector.

An encoding apparatus according to a sixth aspect of the present invention is an adaptive excitation codebook for storing an adaptive excitation vector which is information of an excitation signal of a preceding frame, and pitch periods in each adaptive excitation vector in the adaptive excitation codebook. Pitch position extracting means for extracting a plurality of feature points arranged at intervals as pitch positions, an adaptive excitation vector that minimizes distortion between the decoded speech and the input speech generated with the adaptive excitation vector as the excitation signal, and the adaptive excitation vector An adaptive excitation coding means for obtaining a gain and encoding the adaptive excitation vector and the gain for the adaptive excitation vector; a driving excitation codebook storing a plurality of prepared excitation signals as driving excitation vectors; The pitch position obtained by the pitch position extraction means is combined with the predetermined position of each driving excitation vector of the above driving excitation codebook. And a pitch synchronization means for generating a pitch-synchronized drive excitation vector by repeating the drive excitation vector in a pitch cycle, and a decoded voice generated as a excitation signal a linear sum of the adaptive excitation vector and the pitch-synchronized drive excitation vector. A pitch-synchronized driving sound source vector that minimizes distortion with the input speech and a gain for the pitch-synchronized driving sound source vector are obtained, and the driving sound source vector corresponding to the pitch-synchronized driving sound source vector and the gain for the pitch-synchronous driving sound source vector are encoded. And a driving excitation encoding means for performing the same.

A decoding device according to a seventh aspect of the present invention decodes an adaptive excitation codebook that stores an adaptive excitation vector that is information of an excitation signal of a preceding frame, an adaptive excitation vector and a gain for the adaptive excitation vector. Adaptive excitation decoding means, pitch position extraction means for extracting a plurality of feature points arranged at pitch cycle intervals in the decoded adaptive excitation vector as pitch positions, and storing a plurality of prepared excitation signals as driving excitation vectors Driving excitation codebook and the pitch position obtained by the pitch position extraction means are aligned with a predetermined position of the decoded driving excitation vector, and the driving excitation vector is repeated in a pitch cycle. A pitch synchronization means for generating a vector, a driving sound source vector, and a gay sound source for the pitch synchronization driving sound source vector. Driving excitation decoding means for decoding the above, an adaptive excitation vector, a gain for the adaptive excitation vector, and an excitation vector generation for decoding the excitation vector from the pitch-synchronized excitation vector and the gain for the pitch-synchronized excitation vector And means.

In the encoding / decoding method according to the invention of claim 8, the transmitting side has a first adaptive excitation codebook for storing an adaptive excitation vector which is information of the excitation signal of the preceding frame.
Further, it has a corresponding second adaptive excitation codebook on the receiving side, has a first driving excitation codebook that stores a plurality of prepared excitation signals as driving excitation vectors on the transmitting side, and also has a corresponding first excitation codebook. The receiving side has the second driving excitation codebook, and the transmitting side selects the optimum adaptive excitation vector from the first adaptive excitation codebook for the input speech, determines its gain, and determines the input speech. Alternatively, a plurality of pitch positions are extracted from the adaptive excitation vector, a predetermined position of the driving excitation vector of the driving excitation codebook is aligned with the pitch position to create a pitch-synchronized driving excitation vector, and optimal pitch synchronization is created. A driving sound source vector is selected and its gain is determined, and the adaptive sound source vector, the pitch synchronization driving sound source vector, the adaptive sound source vector, and the pitch synchronization driving sound source vector. The gains for and are encoded and transmitted, and on the receiving side, an adaptive excitation vector is generated from the second adaptive excitation codebook by the reception code, and the gain of the adaptive excitation vector is decoded to obtain the reception code or the adaptation. A plurality of pitch positions are extracted from the sound source vector, and the second
To generate a driving excitation vector from the driving excitation codebook, to create a pitch-synchronized driving excitation vector that is periodic with the predetermined position of the driving excitation vector aligned with the pitch position,
Further, the gain of the pitch-synchronized driving excitation vector is decoded, and the excitation vector is decoded from the adaptive excitation vector, the pitch-synchronized driving excitation vector and the respective gains.

An encoding apparatus according to a ninth aspect of the present invention inputs an adaptive excitation codebook that stores an adaptive excitation vector, which is information on an excitation signal of a preceding frame, and a decoded speech generated by using the adaptive excitation vector as an excitation signal. An adaptive excitation vector that minimizes distortion with speech and an adaptive excitation vector that obtains a gain for the adaptive excitation vector, and encodes the adaptive excitation vector and the gain for the adaptive excitation vector, and a plurality of prepared excitation sources A plurality of pre-prepared driving excitation codebooks that store signals as driving excitation vectors, and gain quantization tables used for driving excitation vectors based on the adaptive excitation signal power information obtained from the adaptive excitation vectors and their gains. Control means for selecting and specifying from the gain quantization table of The driving excitation vector that minimizes the distortion between the decoded speech and the input speech generated by using the linear sum of the driving excitation vector as the excitation signal and the gain for the driving excitation vector are obtained, and the gain quantization table specified above is used. , And a drive excitation encoding means for encoding the drive excitation vector and a gain for the drive excitation vector.

A decoding apparatus according to the invention of claim 10 decodes an adaptive excitation codebook for storing an adaptive excitation vector which is information of an excitation signal of a preceding frame, an adaptive excitation vector and a gain for the adaptive excitation vector. Adaptive excitation decoding means, a driving excitation codebook in which a plurality of excitation signals prepared in advance are stored as driving excitation vectors, power information of the adaptive excitation signal obtained from the adaptive excitation vector and a gain for the adaptive excitation vector By using a control means for selecting and designating a gain quantization table to be used for a driving excitation vector from a plurality of gain quantization tables prepared in advance, and a driving quantization vector specified above. And a driving excitation decoding means for decoding the gain for the driving excitation vector.

The encoding / decoding method according to the invention of claim 11 has, on the transmitting side, a first adaptive excitation codebook for storing an adaptive excitation vector which is information of the excitation signal of the preceding frame.
Further, it has a corresponding second adaptive excitation codebook on the receiving side, has a first driving excitation codebook that stores a plurality of prepared excitation signals as driving excitation vectors on the transmitting side, and also has a corresponding first excitation codebook. The receiving side has two driving excitation codebooks, and the transmitting side selects the optimum adaptive excitation vector from the first adaptive excitation codebook for the input voice, determines the gain, and The gain quantization table of the driving excitation vector is changed by the power information of the adaptive excitation signal obtained from the excitation vector and its gain, and the optimum driving excitation vector is selected from the first driving excitation codebook,
Further, the gain is determined, the gains for the adaptive excitation vector, the driving excitation vector, and the adaptive excitation vector and the driving excitation vector are encoded and transmitted, and on the receiving side, the second adaptive excitation by the reception code. Generating an adaptive excitation vector from the codebook, decoding the gain of the adaptive excitation vector, changing the gain quantization table of the driving excitation vector from the power information of the adaptive excitation signal obtained from the adaptive excitation vector and the gain, Second above
The driving excitation vector is generated from the driving excitation vector codebook, and the gain of the driving excitation vector is decoded, and the excitation vector is decoded from the adaptive excitation vector, the driving excitation vector and each gain. is there.

[0029]

In the input speech coding apparatus according to the present invention, the periodic pulse excitation vector is newly generated corresponding to the adaptive excitation vector, multiplied by each gain and synthesized, and the vector and the gain that minimize distortion with the input speech are obtained. Selected and encoded. Further, the speech decoding apparatus according to the second aspect of the present invention generates an excitation vector having periodicity by the transmitted code and gain, and by this and the adaptive excitation vector,
The source vector is decoded. In the speech coding apparatus according to the third and fourth aspects of the invention, the driving excitation vector is periodically repeated, and the vector and gain that minimize distortion with the input speech are selected and coded. Also, in the speech decoding apparatus according to the fifth aspect of the present invention, the driving excitation vector is periodically generated by the transmitted code, and the excitation vector is decoded by this and the adaptive excitation vector. Further, in the speech coding apparatus according to the invention of claim 6, a feature point determined by a pitch cycle in the adaptive excitation vector is extracted as a pitch position, whereby the driving excitation vector is periodically repeated, and distortion with the input speech is minimized. The vector and gain are selected and encoded. According to the speech decoding apparatus of the invention of claim 7, the pitch position is extracted by the transmitted code, the reception spectrum parameter and the gain, whereby the driving excitation vector is periodically generated, and by this and the adaptive excitation vector. , The excitation vector is decoded.

According to the speech coding / decoding method of the present invention, the pitch is extracted from the adaptive excitation vector in the frame, the pitch synchronization vector synchronized with this pitch is generated and coded, and the adaptive excitation is generated. It is transmitted together with the vector code, and on the receiving side, a pitch synchronization vector which is synchronized with the pitch is generated and added with the adaptive excitation vector to decode the excitation vector. Further, according to the speech coding apparatus of the invention of claim 9, the power information of the adaptive excitation signal is extracted from the adaptive excitation vector and its gain, the gain quantization table for the driving excitation vector is changed in accordance with this, and the input speech The vector and gain that minimize the distortion of are selected and encoded. Further, according to the speech decoding apparatus of the invention of claim 10, the power information of the adaptive excitation signal is extracted by the code of the transmitted adaptive excitation vector and the gain thereof, and the gain quantization table for the driving excitation vector is correspondingly extracted. Modified, which causes the gain for the driving source vector to be decoded and used to decode the source vector. According to the speech encoding / decoding method of the invention of claim 11, the power information of the adaptive excitation signal is extracted from the adaptive excitation vector and its gain, and the gain quantization table for the driving excitation vector is changed in accordance with this, and the input The vector and gain that minimize distortion with speech are selected, coded, and transmitted, and on the receiving side, the gain quantization table for the driving excitation vector is changed according to the power information of the adaptive excitation signal. The gain for the excitation vector is decoded, and this is used to decode the excitation vector.

[0031]

【Example】

Example 1. 1 is a block diagram of an embodiment of claim 1 of the present invention. In FIG. 1, the same parts as those in FIG. 10 are designated by the same reference numerals and the description thereof will be omitted. In FIG. 1, a new part is 20 multi-vector adaptive excitation coding means,
21 and 23 are periodic pulse excitation generating means, and 22 are multiple vector adaptive excitation decoding means. The vocal tract portion including the spectrum analysis unit 6 and the spectrum parameter encoding unit 7 and the vocal cord portion including the multiple vector adaptive excitation encoding unit 20, the driving excitation encoding unit 10, and the excitation vector generation unit 12 are shared as in the conventional example. However, the pitch periodicity of the vocal cords, in particular, is improved, and the voice quality after synthesis is improved.

The operation of the embodiment of the present invention will be described below. First, the operation of the encoding unit 1 will be described. The periodic pulse excitation generator 21 is based on the adaptive excitation vector a _i input from the adaptive excitation codebook 9 and is a periodic pulse excitation vector consisting of a periodic unit impulse sequence in which the dimension having the maximum amplitude is 1 and the other dimensions are 0. generate p _i ,
This p _i is output to the multiple vector coding means 20. Originally, the adaptive excitation codebook 9 stores the sampling sequence value of the input up to the previous frame, and from this, the multi-vector adaptive excitation coding means 20 selects the vector that produces the synthesized speech closest to the input speech of the current frame. However, the stored vector of the adaptive excitation codebook 9 is not sufficient to maintain the periodicity. The periodicity is maintained by newly creating the synchronous pulse source vector source vector p _i .

FIG. 2 shows an example of the adaptive excitation vector (a) and the periodic pulse excitation vector (b) generated from it in this embodiment. That is, of the many stored vectors, the vector with the largest amplitude is unitized and output. This makes it possible to emphasize the periodicity that is important for audio reproduction.

The multi-vector adaptive excitation coding means 20 is
Using the spectrum parameters input from the spectrum parameter encoding means 7, the adaptive excitation vector a _i input from the adaptive excitation codebook 9 and the periodic pulse excitation vector p _i input from the periodic pulse excitation generation means 21.
More synthesized speech vectors A _i and P _i are synthesized respectively. Then, the inter-vector distance D _i with respect to the input voice vector X cut out from the input voice 4 for each frame is calculated by, for example, formula (5)
Ask according to. Here, the gains β _Ai and β _Pi are distances D _i
Is determined according to, for example, equations (6) and (7).

[0035]

[Equation 3]

Next, a set of vectors A _I and P _I that minimizes the inter-vector distance is searched, and the code I and the gains β _qAI and β _{qPI obtained} by quantizing the gains β _AI and β _PI are transmitted through the transmission line. 3 to the decoding unit 2. Also, the selected adaptive source vector a _I and its gain β _qAI , the periodic pulse source vector p _I and its gain β _qPI are output to the source vector generation means 12, and the error vector X is also output.
As (2), X * = X−β _qAI A _I −β _qPI P _I is output to the driving _excitation encoding means 10. Thus, as compared with the conventional, it is novel that the vector of A _I and the set of vectors of A _I and P _I. The operation of the driving _excitation coding means 10 for further quantizing and coding and transmitting the vector difference is the same as the conventional example in this embodiment, and the codes of J and γ _qJ are transmitted to the decoding unit 2. To do.

Next, the operation of the decoding section 2 will be described. The multiple vector adaptive excitation decoding means 22 outputs the code I of the adaptive excitation vector input from the encoding unit 1 to the adaptive excitation codebook 14. The adaptive excitation codebook 14 converts the adaptive excitation vector a _I corresponding to the code I into multiple vector adaptive excitation decoding means 22 and periodic pulse excitation generation means 23.
Output to. The periodic pulse excitation generating means 23 generates a periodic pulse excitation vector p _I from the adaptive excitation vector a _I input from the adaptive excitation codebook 14 and outputs it to the multi-vector adaptive excitation decoding means 22.

The multi-vector adaptive excitation decoding means 22 is
From the adaptive excitation vector a _I input from the adaptive excitation codebook 14, the periodic pulse excitation vector p _I input from the periodic pulse excitation generation means 23, and the respective gain codes input from the encoding unit 1. The decoded gains β _qAI and β _qPI are output to the _excitation vector generation means 17.
That is, a set of β _qAI , a _I , β _qPI and p _I is given to the sound source vector generating means.

The excitation vector generating means 17 receives the adaptive _excitation vector a _I , the periodic pulse _excitation vector p _I and the respective gains β _qAI and β _qPI inputted from the multi-vector adaptive _excitation decoding means 22 and the other input. The following _excitation vector is generated from the driving _excitation vector c _J input from a certain driving _excitation decoding means 15 and its gain γ _qJ . β _qAI a _I + β _qPI p _I + γ _qJ c _{J The} generated _excitation vector is output to the adaptive _excitation codebook 14 and the synthesis filter 19. The output of the synthesizing filter 19 is a voice output that is synthesized and restored.

Example 2. In the first embodiment, the periodic pulse excitation vector is generated by considering only the pitch periodicity in the frame to be encoded. On the other hand, the pitch period of the adaptive excitation vector that is cut out and selected / encoded has a value that is twice the actual pitch period of the input speech, and the pulse train generated by the periodic pulse excitation vector is not necessarily periodic between frames. It doesn't. At this time, using the information of the preceding frame, for example, the average pitch period is obtained in advance, and the periodic pulse excitation vector is generated from the pulse train of the average pitch period. A pulsed source vector may be generated.

Example 3. In the first embodiment, the amplitude of each pulse of the periodic pulse source vector is assumed to be equal. By changing this, for example, the amplitude is changed linearly by sequentially increasing, decreasing, or the like from the previous frame according to the power of the voice, and setting so that each periodic pulse source vector has a different amplitude. Good.

Example 4. In the first embodiment, the pulse position of the periodic pulse excitation vector is obtained from the maximum amplitude point of the adaptive excitation vector. In the synthesized speech vector obtained by synthesizing the two, the inner product of the two vectors becomes maximum. A periodic pulse source vector may be set.

Example 5. In the first embodiment, the adaptive sound source vector and the periodic pulse sound source vector are used as they are to generate a synthetic speech vector, and the inter-vector distance between the resultant vector and the input speech vector is obtained. However, for example, the amplitude value of the adaptive excitation vector at the pulse position of the periodic pulse excitation vector is set to 0, the adaptive excitation vector and the periodic pulse excitation vector are orthogonalized, and the correlation between the vectors is removed. You may generate the synthetic | combination speech vector using. Then, the inter-vector distance between the result vector and the input voice vector may be obtained.

Example 6. In the fifth embodiment, the adaptive excitation vector and the periodic pulse excitation vector are orthogonalized to remove their correlation. The correlation between the vectors may be removed by orthogonalizing the synthesized speech vectors obtained by synthesizing them.

Example 7. In the first embodiment, the periodic pulse sound source vector is composed of one vector. In the case where the gain of each pulse of the periodic pulse source vector is two pulses in FIG. 2B, the gain of the previous pulse and the gain of the subsequent pulse may be set to different values in terms of time. .

Example 8. FIG. 3 is a block diagram showing an embodiment of the inventions of claims 3 and 5 of the present invention. In FIG. 3, the same parts as those in FIG. 1 are designated by the same reference numerals and the description thereof will be omitted. In FIG. 3, 24 is an adaptive excitation coding means, 25 is a pitch position extraction means, and 26 is a driving excitation coding means. Reference numerals 27 and 28 are pitch synchronization means, and 29 is a sound source vector generation means.

The operation of the embodiment of the present invention shown in FIG. 3 will be described below. First, the operation of the encoding unit 1 will be described. The adaptive excitation coding means 24 synthesizes the synthesized speech vector A _i using the adaptive excitation vector a _i input from the adaptive excitation codebook 9 and the spectrum parameter input from the spectrum parameter coding means 7. Then, a synthetic speech vector A _I and its gain β _I that minimize distortion with the input speech vector X cut out from the input speech 4 for each frame are searched, and its code I and gain β _I are searched.
The code of the gain β _qI quantized is output to the decoding unit 2 via the transmission line 3, and the selected adaptive _excitation vector a _I and its gain β _qI are output to the excitation vector generation means 12,
The addition error vector X ■, X ■ = pitch is X-β _qI A _I, and outputs the X ■ and the adaptive excitation vector a _I to driving excitation coding unit 26, corresponding to the adaptive excitation vector a _I Pitch position extraction means 2
Output to 5.

The pitch position extraction means 25 creates a pulse train of a pitch cycle using the pitch cycle input from the adaptive excitation coding means 24, and uses this as a sound source for the spectrum parameter input from the spectrum parameter coding means 7. When a synthetic speech vector is generated using the input speech vector X, the input speech vector X is extracted from the input speech 4 for each frame.
Search for a pulse train that minimizes the distortion between and. Then, using the pulse position as the pitch position, the pitch synchronization means 27
Output to.

FIG. 4 shows an operation block diagram when the pitch position is searched. In this operation, the start position of the pulse of the pitch cycle is divided from the start position of the frame, for example, by dividing the cut-out time of one frame into 128 equal parts, and a value shifted by 1/128 hours is sequentially set. Then, the one obtained by combining the pulse trains whose pitch positions are shifted and closest to the input voice is selected. By doing so, a point that is uniquely determined for the pitch structure of the voice is required regardless of the framing of the input voice, and thus the periodicity can be clarified.

In the driving excitation codebook 11, for example, as a result of learning by the Linde, Buzo, Gray method (LBG method) from the short cycle prediction residual signal cut out at a pitch length synchronized with the pitch position of the speech, N pieces are obtained. The drive vector of is stored. In addition, the pitch position where each driving sound source vector is cut out is set as the pitch synchronization position of the driving sound source vector. The pitch synchronization means 27 aligns the pitch synchronization position of the drive excitation vector c _j input from the drive excitation codebook 11 with the pitch position input from the pitch position extraction means 25, and repeats the pitch synchronization drive in the pitch cycle. A sound source vector c _j ■ is created. Then, this is output to the driving excitation encoding means 26. FIG. 5 shows an example of the pitch synchronous drive sound source vector in this embodiment. The driving excitation coding means 26 converts the pitch-synchronized driving excitation vector c _{j 2} input from the pitch synchronization means 27 into the adaptive excitation vector a _I input from the adaptive excitation coding means 24.
On the other hand, for example, the orthogonalized pitch-synchronized drive sound source vector c _{j {} circle around (2) _} is created according to equation (8).

[0051]

[Equation 4]

Next, a synthesized voice vector C _{j {1} } is synthesized from the orthogonal pitch-synchronized drive sound source vector c _{j {1} }. Then, the inter-vector distance D _j from the error vector X ■ input from the adaptive excitation encoding means 24 is calculated by, for example, equation (9).
Ask according to. Here, the gain γ _j is determined, for example, according to the equation (10) so that the inter-vector distance D _j is minimized.

[0053]

[Equation 5]

Next, a vector C _{J (1)} that minimizes the inter-vector distance D _j is searched for, and the code J and the gain γ _J are searched.
The quantized γ _qJ code is output to the decoding unit 2 via the transmission line 3. At the same time, the selected orthogonalized pitch-synchronized drive sound source vector c _{J (1)} and its gain γ _qJ are output to the sound source vector generating means 12. This improves the periodicity of the sound source signal by synchronizing the driving sound source vector, which has been disturbing the periodicity of the sound source signal generated conventionally because it is asynchronous with the pitch, with the pitch.

Next, the operation of the decoding section 2 will be described. The pitch synchronization means 28 creates a pitch-synchronized drive excitation vector c _J (1) that is repeated in synchronization with the pitch position input from the encoder 1 from the drive excitation vector c _J input from the drive excitation codebook 16. , To the driving excitation decoding means 15.

The excitation vector generating means 29 first orthogonalizes the pitch-synchronized driving excitation vector c _{J (2)} input from the driving excitation decoding means 15 with the adaptive excitation vector a _I input from the adaptive excitation decoding means 13. To generate an orthogonalized pitch-synchronized drive sound source vector c _J ∘. Next, the gain β _qI input from the adaptive _excitation vector a _I and the adaptive _excitation decoding means 13, and the gain γ _qJ input from the orthogonalized pitch-synchronized drive _excitation vector c _{J 1} and the drive _excitation decoding means 15. From this, an _excitation vector β _qI a _I + γ _qJ c _J 2 of the following equation is generated and output to the adaptive _excitation codebook 14 and the synthesis filter 19. The output of the synthesizing filter 19 becomes an audio output that is synthesized and restored.

Embodiment 9 FIG. 6 is a block diagram showing an embodiment of the inventions of claims 4 and 5 of the present invention. 6, parts that are the same as those in FIG. 3 are given the same reference numerals, and descriptions thereof will be omitted. In FIG. 6, 30 is a drive excitation encoding means.

An embodiment of the present invention shown in FIG. 6 will be described below. The adaptive excitation coding means 24 synthesizes the synthesized speech vector A _i using the adaptive excitation vector a _i input from the adaptive excitation codebook 9 and the spectrum parameter input from the spectrum parameter coding means 7.
Then, from the input speech 4 distortion of the input speech vector X cut out for each frame is searched vector A _I and gain beta _I minimized, of the code I, and the gain beta _qI the gain beta _I quantized The code is output to the decoding unit 2 via the transmission path 3, the selected adaptive _excitation vector a _I and its gain β _qI are output to the excitation vector generation means 12, and the error vector X ■ is _expressed as X ■ = X. −β _qI A _I, and outputs the pitch period corresponding to this X 2 and the adaptive _excitation vectors a _I and a _I to the driving excitation encoding means 30.

The driving excitation coding means 30 first outputs the pitch position k to the pitch synchronization means 27 using the pitch period input from the adaptive excitation coding means 24. The pitch position k is a position determined by the pitch cycle starting from the k-th position from the head position of the frame. Pitch synchronization means 2
Reference numeral 7 is a pitch-synchronized drive excitation vector which is repeated at pitch intervals by matching the pitch synchronization position of the drive excitation vector c _j input from the drive excitation codebook 11 with the pitch position k input from the drive excitation encoding means 30. Create c _jk ■. Then, this is output to the driving excitation encoding means 30. The driving excitation coding means 30 is a pitch-synchronized driving _excitation vector c _{jk (2)} input from the pitch synchronization means 27.
Then, the orthogonalized pitch-synchronized drive _excitation vector c _jk (2) is created by orthogonalizing the adaptive excitation vector a _I input from the adaptive excitation encoding means 24 according to, for example, equation (11).

[0060]

[Equation 6]

Next, a synthesized speech vector C _{jk s} is synthesized from the orthogonalized pitch synchronization drive sound source vector c _{jk s} .
Then, the inter-vector distance D _jk with the error vector X2 input from the adaptive excitation coding means 24 is obtained according to, for example, equation (12). Here, the gain γ _jk is determined, for example, according to the equation (13) so that the inter-vector distance D _jk is minimized.

[0062]

[Equation 7]

Next, the inter-vector distance D _{jk is set} to the code j (j = 1, ..., N) of the driving sound source vector and the pitch position k.
For all combinations (k = 1, ..., L), a vector C _JK (2) that minimizes the inter-vector distance D _jk is searched, and this code J, pitch position K, and gain γ _JK are obtained. The quantized γ _qJK code is decoded by the decoding unit 2 via the transmission line 3.
Output to. At the same time, the selected orthogonalized pitch-synchronized drive sound source vector c _{JK 1} and its gain γ _qJK are output to the sound source vector generating means 12. This improves the sound quality of the synthetic voice by extracting the pitch position as the one that minimizes the distortion of the synthetic voice.

Example 10. FIG. 7 is a block diagram showing an embodiment of the invention of claim 6 of the present invention. 7, parts that are the same as those shown in FIG. 6 are given the same reference numerals, and descriptions thereof will be omitted. In FIG. 7, 31 is adaptive excitation coding means, 3
2 and 33 are pitch position extracting means.

The operation of the embodiment of the present invention shown in FIG. 7 will be described below. First, the operation of the encoding unit 1 will be described. The pitch position extracting means 32 synthesizes the synthesized speech vector A _i using the adaptive excitation vector a _i input from the adaptive excitation codebook 9 and the spectrum parameter input from the spectrum parameter encoding means 7. Next, when a synthesized speech vector is generated using the spectral parameters with a pulse train having a pitch period corresponding to the adaptive source vector a _i as a sound source, a pulse train having a minimum distortion with the synthesized speech vector A _i is searched for. To do. Then, the pulse position is output as a pitch position to the adaptive excitation encoding means 31.

The adaptive excitation coding means 31 first uses the pulse train of the pitch period corresponding to the adaptive excitation vector a _i input from the adaptive excitation codebook 9 as the excitation, and inputs the spectrum input from the spectrum parameter encoding means 7. When a synthetic speech vector is generated using the parameters, a pulse train that minimizes distortion with the input speech vector X cut out from the input speech 4 for each frame is searched for. Then, the pulse position is set as the pitch position in the input voice. Then, the following processing is performed only when the pitch position in the adaptive sound source vector a _i input from the pitch position extraction means 32 is within a certain distance, for example, 5 samples, from the pitch position in the input speech, and other than that. In this case, the adaptive excitation vector a _i is excluded from the target of encoding.

The adaptive excitation coding means 31 then generates a pitch pulse vector p _{i in} which only the pitch position in the adaptive excitation vector a _i is 1 and the others are 0. Then, using the spectrum parameters, synthesized speech vectors A _i and P _i are synthesized from the adaptive sound source vector a _i and the pitch pulse vector p _i . Further, the inter-vector distance D _i with respect to the input voice vector X is obtained, for example, according to the equation (14). Here, min (x, y) is a function that selects the smaller one of x and y, and the selection result is output.
Further, the gains β _Ai and β _Pi are determined in accordance with, for example, equations (15) and (16) so that the distance D _i becomes the minimum.

[0068]

[Equation 8]

Next, the vector A _I having the minimum inter-vector distance is searched for, and the code I and the code of the gain β _{qAI obtained} by quantizing the gain β _AI are transmitted to the decoding unit 2 via the transmission line 3. Output. Further, the adaptive _excitation vector a _I and its gain β _qAI are output to the excitation vector generation means 12. Further, the pitch position in the adaptive sound source vector a _I input from the pitch position extraction means 32 is output to the pitch synchronization means 27. Also, the error vector X
X X = X-β _qAI A _I, and _outputs this X ■ and the adaptive _excitation vector a _I to the driving excitation encoding means 26.

Next, the operation of the decoding section 2 will be described. The pitch position extracting means 33 extracts a pitch position from the adaptive excitation vector a _I input from the adaptive excitation codebook 14 and the spectrum parameter input from the spectrum parameter decoding means 18. Then, this pitch position is output to the pitch synchronizing means 28. This can be uniquely obtained in the decoding unit by extracting the pitch position from the adaptive excitation vector that is a quantized input speech and the spectrum parameter, without the need to transmit the pitch position, The amount of transmitted information can be reduced without deteriorating the sound quality of the synthetic voice.

Example 11. In the tenth embodiment, the adaptive excitation encoding means fixes the distance value between the pitch position extracted from the adaptive excitation vector which is the criterion for determining the adaptive excitation vector as the encoding target and the pitch position obtained from the input speech. However, it may be variable according to the pitch length, for example.

Example 12 In the tenth embodiment, the adaptive excitation coding means targets the encoding only when the difference between the pitch position extracted from the adaptive excitation vector and the pitch position obtained from the input speech is within a certain distance. This distance criterion may be relaxed or eliminated in frames where the vector does not meet this criterion.

Example 13 FIG. 8 is a block diagram showing an embodiment of the invention of claim 9 of the present invention. 8, parts that are the same as the parts shown in FIG. 10 are given the same reference numerals, and descriptions thereof will be omitted. In FIG. 8, 34 and 36 are control means, and 35.
Is a driving excitation encoding means, and 37 is a driving excitation decoding means.

The operation of the embodiment of the present invention shown in FIG. 8 will be described below. First, the operation of the encoding unit 1 will be described. The control means 34 _obtains the power P P = ‖β _qI a _I ‖ ² of the adaptive _excitation signal from the adaptive _excitation vector a _I input from the adaptive _excitation encoding means 8 and its gain β _qI . Then, for example, a preset threshold P ₁ ,
Comparing P ₂ and P ₃ with the power P of the adaptive sound source signal,
From the gain quantization table for the driving excitation vector prepared in advance, for example, select the gain quantization table for the driving excitation vector to be actually used for quantization according to the following condition determination, and select the selected quantization table. It outputs to the drive excitation encoding means 35. If P <P ₁ Select Quantization Table 1 If P ₁ ≦ P <P ₂ Select Quantization Table 2 If P ₂ ≦ P <P ₃ Select Quantization Table 3 If P ₃ ≦ P Quantization Select table 4

Each quantization table is designed so that efficient quantization can be performed under each condition by using the correlation between the power of the adaptive excitation signal and the gain with respect to the driving excitation vector shown in FIG. 9, for example. As can be seen from FIG. 9, the gain range to be quantized in each quantization table can be reduced by changing the quantization of the gain for the driving excitation vector according to the power information of the adaptive excitation signal, thus increasing the quantization efficiency. be able to.

The driving excitation coding means 35 synthesizes the synthesized speech vector C _j using the driving excitation vector c _j inputted from the driving excitation codebook 11 and the spectrum parameter inputted from the spectrum parameter coding means. .
Then, the inter-vector distance D _j with the error vector X₄ input from the adaptive excitation encoding means 8 is obtained according to, for example, equation (17). Here, the gain γ _j is determined, for example, according to the equation (18) so that the distance D _j is minimized.

[0077]

[Equation 9]

Next, a vector C _J that minimizes the inter-vector distance is searched, and its code J and gain γ _J are quantized using the gain quantization table for the driving sound source vector input from the control means 34. The code of the gain γ _qJ is output to the decoding unit 2 via the transmission path 3, and the selected drive _excitation vector c _J and its gain γ _qJ are output to the _excitation vector generation means 12.

Next, the operation of the decoding section 2 will be described. The control means 36 _obtains the power P of the adaptive _excitation signal from the adaptive _excitation vector a _I input from the adaptive _excitation decoding means 13 and its gain β _qI, and a plurality of drive _excitation sources, which are the same as the control means 34, are prepared in advance. From the gain quantization table for the vector, the gain quantization table for the driving excitation vector used for inverse quantization is selected according to the condition determination using the same threshold value as that of the control means 34, and the selected quantization table is used as the driving excitation source. It outputs to the decoding means 37.

The driving excitation decoding means 37 reads the driving excitation vector c _J from the driving excitation codebook 16 based on the code J of the driving excitation vector input from the encoding unit 1,
Further, using the gain quantization table input from the control means 36, the gain γ _qJ is decoded from the sign of the gain for the driving _excitation vector input from the encoding unit 1,
The driving sound source vector c _J and its gain γ _qJ are output to the sound source vector generating means 17.

[0081]

As described above, according to the present invention, an adaptive excitation codebook that stores an adaptive excitation vector and a periodic pulse excitation generation means that generates a periodic pulse excitation vector are added to a speech encoding device. Since the multi-vector adaptive excitation encoding means for synthesizing and encoding is provided, and the speech decoding apparatus is provided with means corresponding to the encoding apparatus, only a small amount of transmission information is added, and the pitch periodicity This has the effect of synthesizing high-quality speech with little disturbance.

According to the inventions of claims 3 to 5, the speech coding apparatus comprises an adaptive excitation codebook, pitch position extraction means, a driving excitation codebook storing a plurality of excitation signals, Since the pitch synchronization means for repeating the driving excitation vector in the pitch cycle, the adaptive excitation coding means, and the driving excitation coding means are provided, the speech decoding apparatus is provided with means corresponding to the coding apparatus. Therefore, it is possible to synthesize high-quality speech with little disturbance of pitch periodicity by adding a small amount of transmission information.

According to the sixth and seventh aspects of the invention, the speech coding apparatus comprises an adaptive excitation codebook, pitch position extraction means for extracting a pitch position from an adaptive excitation vector, and a driving excitation codebook. Since the pitch synchronization means for repeating the driving excitation vector in the pitch cycle, the adaptive excitation coding means, and the driving excitation coding means are provided, the speech decoding apparatus is provided with means corresponding to the coding apparatus. So
There is an effect that it is possible to synthesize high-quality speech with little disturbance of pitch periodicity without increasing transmission information.

Further, according to the invention of claim 8, since the method for extracting the periodic pitch and the method for generating the pitch synchronization vector having the periodicity are provided on the receiving side, the pitch can be obtained by adding a small amount of transmission information. This has the effect of enabling high-quality transmission with less periodic disturbance.

According to the ninth and tenth aspects of the present invention, the speech coding apparatus is driven by the adaptive excitation codebook, the driving excitation codebook, the adaptive excitation coding means, and the power information of the adaptive excitation signal. Since the control means for changing the gain quantization table for the excitation vector and the driving excitation encoding means are provided, and the speech decoding apparatus is provided with the means corresponding to the encoding apparatus, there is no increase in transmission information. In addition, there is an effect that high-quality speech with small quantization distortion can be synthesized.

Further, according to the invention of claim 11, a method of changing the quantization of the gain with respect to the driving excitation vector to an efficient one by the power information of the adaptive excitation signal, and a quantization corresponding to the receiving side and the transmitting side. Since there is provided a method of changing to, there is an effect that high-quality transmission with small quantization distortion can be performed without increasing the transmission information.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a first embodiment of the present invention.

FIG. 2 is an explanatory diagram showing an example of a periodic pulse sound source vector of the present invention.

FIG. 3 is a configuration diagram showing an eighth embodiment of the present invention.

FIG. 4 is an explanatory diagram showing pitch position extraction of the present invention.

FIG. 5 is an explanatory diagram showing an example of a pitch-synchronized drive sound source vector of the present invention.

FIG. 6 is a configuration diagram showing an embodiment 9 of the present invention.

FIG. 7 is a configuration diagram showing an embodiment 10 of the present invention.

FIG. 8 is a configuration diagram showing an embodiment 13 of the present invention.

FIG. 9 is an explanatory diagram showing the relationship between the power of the adaptive excitation signal and the gain with respect to the driving excitation vector.

FIG. 10 is a configuration diagram showing a conventional speech encoding / decoding device.

FIG. 11 is an explanatory diagram showing an example of an adaptive sound source vector.

[Explanation of symbols]

 1 Encoding Unit 2 Decoding Unit 3 Transmission Line 4 Input Speech 5 Output Speech 6 Spectrum Analysis Means 7 Spectrum Parameter Encoding Means 8 Adaptive Excitation Encoding Means 9, 14 Adaptive Excitation Codebook 10 Driving Excitation Encoding Means 11, 16 Driving Excitation codebook 12, 17 Excitation vector generation means 13 Adaptive excitation decoding means 15 Drive excitation decoding means 18 Spectral parameter decoding means 19 Synthesis filter 20 Multiple vector adaptive excitation coding means 21, 23 Periodic pulse excitation generation means 22 Multiple vectors Adaptive excitation decoding means 24 Adaptive excitation coding means 25 Pitch position extraction means 26 Drive excitation coding means 27, 28 Pitch synchronization means 29 Excitation vector generation means 30 Drive excitation coding means 31 Adaptive excitation coding means 32, 33 Pitch Position extraction means 34, 36 Control means 35 Drive sound Coding means 37 driving excitation decoding means

Claims (11)

[Claims] 1. A speech coder which divides an input speech into spectral envelope information and excitation signal information for each frame divided at regular intervals, and stores an adaptive excitation vector, which is information on the excitation signal of a preceding frame. Adaptive excitation codebook, a periodic pulse excitation generating means for generating a periodic pulse excitation vector determined as a main component of each adaptive excitation vector in the adaptive excitation codebook, the adaptive excitation vector and a period corresponding to the adaptive excitation vector An adaptive excitation vector that minimizes distortion between the decoded speech and the input speech generated by using a linear sum of the pulse excitation vector as the excitation signal, a periodic pulse excitation vector corresponding to the adaptive excitation vector, the adaptive excitation vector, and the periodic pulse The gain with respect to the sound source vector is obtained, and the adaptive sound source vector and the adaptive sound source vector A speech coding apparatus comprising a plurality of vector adaptive excitation coding means for coding a gain for the periodic pulse excitation vector. 2. A speech decoding apparatus which decodes spectrum envelope information and excitation signal information coded for each frame and generates decoded speech using each decoded parameter, wherein the excitation signal of the preceding frame is An adaptive excitation codebook that stores an adaptive excitation vector that is information, a periodic pulse excitation generation means that generates a periodic pulse excitation vector that is determined as a main component of the decoded adaptive excitation vector, an adaptive excitation vector, and the adaptive excitation vector And a multi-vector adaptive excitation decoding means for decoding gains for the periodic pulse excitation vector corresponding to the adaptive excitation vector and for adding the gains to the adaptive excitation vector and the periodic pulse excitation vector respectively. Speech decoding device. 3. A speech coding apparatus for coding an input speech by dividing it into spectral envelope information and excitation signal information for each frame divided at regular intervals, and storing an adaptive excitation vector which is information of excitation signal of a preceding frame. Adaptive excitation codebook, an adaptive excitation vector that minimizes distortion between the decoded speech generated by using the adaptive excitation vector as the excitation signal and the input speech, and a gain for the adaptive excitation vector are obtained, and the adaptive excitation vector and the adaptive excitation vector Adaptive excitation coding means for coding a gain for a vector, a driving excitation codebook in which a plurality of prepared excitation signals are stored as driving excitation vectors, and a plurality of features arranged at pitch cycle intervals with respect to input speech A pitch position extracting means for extracting a point as a pitch position and encoding the pitch position; Pitch synchronization means for aligning a predetermined position of each driving excitation vector of the dynamic excitation codebook and generating a pitch synchronization driving excitation vector by repeating the driving excitation vector in a pitch cycle; the adaptive excitation vector and the pitch synchronization driving excitation. Pitch-synchronized driving sound source vector that minimizes distortion between the decoded speech generated by linearly summing the vector as a sound source signal and the input speech, and the gain for the pitch-synchronized driving sound source vector, and driving corresponding to the pitch-synchronized driving sound source vector A speech encoding apparatus comprising an excitation vector and a drive excitation encoding means for encoding a gain for the pitch-synchronized drive excitation vector. 4. A speech coding apparatus for coding an input speech by dividing it into spectral envelope information and excitation signal information for each frame divided at regular intervals, and storing an adaptive excitation vector which is information of excitation signal of a preceding frame. Adaptive excitation codebook, an adaptive excitation vector that minimizes distortion between the decoded speech generated by excitation signal of the adaptive excitation vector and the input speech, and a gain for the adaptive excitation vector are obtained, and the adaptive excitation vector and the adaptive An adaptive excitation coding means for encoding a gain for an excitation vector, a driving excitation codebook that stores a plurality of prepared excitation signals as driving excitation vectors, and a plurality of points arranged at pitch cycle intervals as pitch positions. , Aligning a predetermined position of each driving excitation vector of the driving excitation codebook with the pitch position, Pitch synchronization means for repeatedly generating a pitch-synchronized driving sound source vector in a pitch cycle, and distortion of a decoded speech and an input speech generated by using a linear sum of the adaptive sound source vector and the pitch-synchronized driving sound source vector as a sound source signal is minimized. A drive excitation code for obtaining a pitch-synchronized drive excitation vector and a gain for the pitch-synchronized drive excitation vector, and encoding a drive excitation vector corresponding to the pitch-synchronized drive excitation vector, a pitch position, and a gain for the pitch-synchronized drive excitation vector. A speech coding apparatus having a coding means. 5. A speech decoding apparatus for decoding spectral envelope information and excitation signal information encoded for each frame and generating decoded speech using each of the decoded parameters, wherein the excitation signal of the preceding frame is An adaptive excitation codebook that stores an adaptive excitation vector that is information, an adaptive excitation decoding unit that decodes an adaptive excitation vector and a gain for the adaptive excitation vector, and a plurality of prepared excitation signals that are stored as driving excitation vectors The driving excitation codebook that is operating, the pitch position is decoded, the predetermined position of the decoded driving excitation vector is aligned with the above pitch position, and the above driving excitation vector is repeated in a pitch cycle to generate a pitch synchronization driving excitation vector. Pitch synchronization means, driving source vector, and gain for the pitch-synchronized driving source vector Driving excitation decoding means for performing the above, and the excitation vector generating means for decoding the excitation vector from the adaptive excitation vector and the gain for the adaptive excitation vector, and the gains for the pitch-synchronized excitation vector and the pitch-synchronized excitation vector. A provided voice decoding device. 6. A speech coding apparatus for coding an input speech by dividing it into spectral envelope information and excitation signal information for each frame divided at regular intervals, and storing an adaptive excitation vector which is information of an excitation signal of a preceding frame. Adaptive excitation codebook, pitch position extracting means for extracting a plurality of feature points arranged at pitch cycle intervals in each adaptive excitation vector in the adaptive excitation codebook as pitch positions, and the adaptive excitation vector as the excitation signal. An adaptive excitation vector that obtains an adaptive excitation vector that minimizes distortion between the decoded speech and the input speech, a gain for the adaptive excitation vector, and an adaptive excitation vector that encodes the adaptive excitation vector and the gain for the adaptive excitation vector, and is prepared in advance. Driving excitation codebook storing a plurality of excitation signals as driving excitation vectors, and the pitch position Pitch synchronization means for aligning a predetermined position of each driving excitation vector of the driving excitation codebook with the pitch position obtained by the outputting means, and generating a pitch synchronization driving excitation vector by repeating the driving excitation vector in a pitch cycle. , A pitch-synchronous excitation vector that minimizes distortion between the decoded speech generated by using a linear sum of the adaptive excitation vector and the pitch-synchronized excitation vector as the excitation signal and the input speech, and a gain for the pitch-synchronous excitation vector A speech coding apparatus comprising: a driving excitation vector corresponding to the pitch-synchronized driving excitation vector; and driving excitation coding means for coding a gain for the pitch-synchronized driving excitation vector. 7. A speech decoding apparatus which decodes spectrum envelope information and excitation signal information encoded for each frame and generates decoded speech using each decoded parameter, in a speech signal of a preceding frame. An adaptive excitation codebook that stores an adaptive excitation vector that is information; an adaptive excitation decoding means that decodes an adaptive excitation vector and a gain for the adaptive excitation vector; and a plurality of decoded adaptive excitation vectors arranged at pitch cycle intervals. Pitch position extracting means for extracting the feature points of as a pitch position, a driving excitation codebook that stores a plurality of prepared excitation signals as driving excitation vectors, and a pitch position obtained by the pitch position extracting means. , A predetermined position of the decoded driving sound source vector is aligned, and the driving sound source vector is repeated in a pitch cycle. Pitch synchronization means for generating a synchronous drive excitation vector, a drive excitation vector, a drive excitation decoding means for decoding a gain for the pitch synchronization drive excitation vector, the adaptive excitation vector and a gain for the adaptive excitation vector A speech decoding apparatus comprising an excitation vector generation means for decoding an excitation vector from the pitch synchronization driven excitation vector and a gain for the pitch synchronization driven excitation vector. 8. A speech coding / decoding method for transmitting input speech by dividing the speech into spectral envelope information and excitation signal information for each frame divided at regular intervals, wherein an adaptive excitation vector which is information of the excitation signal of the preceding frame is set. A first adaptive excitation codebook to be stored is provided on the transmitting side, a corresponding second adaptive excitation codebook is provided on the receiving side, and a plurality of prepared excitation signals are stored as drive excitation vectors. Has the drive excitation codebook of
Further, the corresponding second driving excitation codebook is provided on the receiving side, and the transmitting side selects the optimum adaptive excitation vector from the first adaptive excitation codebook for the input voice and sets its gain. A plurality of pitch positions are extracted from the input speech or the adaptive excitation vector, and a pitch-synchronized drive excitation vector is created by synchronizing the pitch positions with predetermined positions of the drive excitation vector of the drive excitation codebook. An optimum pitch-synchronized driving sound source vector is selected and its gain is determined, and the adaptive sound source vector, the pitch synchronous driving sound source vector, the gains for the adaptive sound source vector and the pitch synchronous driving sound source vector are encoded and transmitted. , The receiving side generates an adaptive excitation vector from the second adaptive excitation codebook by the received code, and Decoding a plurality of pitch positions from the received code or the adaptive excitation vector, generating a driving excitation vector from the second driving excitation codebook, and setting a predetermined position of the driving excitation vector at the pitch position. A voice that creates a pitch-synchronized driving sound source vector that is also cycled, decodes the gain of the pitch-synchronized driving sound source vector, and decodes the sound source vector from the adaptive sound source vector, the pitch-synchronous driving sound source vector, and respective gains Encoding and decoding method. 9. A speech coding apparatus for coding an input speech by dividing it into spectral envelope information and excitation signal information for each frame divided at regular intervals, and storing an adaptive excitation vector which is information of an excitation signal of a preceding frame. Adaptive excitation codebook, an adaptive excitation vector that minimizes distortion between the decoded speech generated by using the adaptive excitation vector as the excitation signal and the input speech, and a gain for the adaptive excitation vector are obtained, and the adaptive excitation vector and the adaptive excitation vector Adaptive excitation coding means for encoding a gain for a vector, a driving excitation codebook in which a plurality of prepared excitation signals are stored as driving excitation vectors, and an adaptive excitation signal obtained from the adaptive excitation vector and its gain The gain quantization table used for the driving sound source vector is prepared in advance according to the power information of Control means for selecting and designating from among a number of gain quantization tables, and driving for minimizing distortion between the decoded speech and the input speech generated by using the linear sum of the adaptive excitation vector and the driving excitation vector as the excitation signal. A speech code including a drive source vector and a drive source encoding unit that encodes a gain for the drive source vector by obtaining a gain for the drive source vector and using the specified gain quantization table. Device. 10. A speech decoding apparatus for decoding spectral envelope information and excitation signal information encoded for each frame and generating decoded speech using each of the decoded parameters, wherein the excitation signal of the preceding frame is An adaptive excitation codebook that stores an adaptive excitation vector that is information, an adaptive excitation decoding unit that decodes an adaptive excitation vector and a gain for the adaptive excitation vector, and a plurality of prepared excitation signals that are stored as driving excitation vectors Driving excitation codebook, and the gain quantization table used for the driving excitation vector based on the adaptive excitation vector and the power information of the adaptive excitation signal obtained from the gain for the adaptive excitation vector. Of the specified gain quantization table Using Bull, speech decoding apparatus and a driving excitation decoding means for decoding the gain excitation vector and to its driving excitation vector. 11. A speech coding / decoding method for transmitting input speech by dividing the speech into spectral envelope information and excitation signal information for each frame divided at regular intervals, wherein an adaptive excitation vector which is information of the excitation signal of a preceding frame is set. A first adaptive excitation codebook to be stored is provided on the transmitting side, a corresponding second adaptive excitation codebook is provided on the receiving side, and a plurality of prepared excitation signals are stored as drive excitation vectors. Has the drive excitation codebook of
The corresponding second driving excitation codebook is provided on the receiving side, and the transmitting side selects the optimum adaptive excitation vector from the first adaptive excitation codebook with respect to the input voice, and sets its gain. The quantization table of the driving excitation vector is changed according to the power information of the adaptive excitation signal obtained from the adaptive excitation vector and its gain, and the optimum driving excitation vector is selected from the first driving excitation codebook. , The gain is determined, the gains for the adaptive excitation vector, the driving excitation vector, and the adaptive excitation vector and the driving excitation vector are encoded and transmitted, and on the receiving side, the second adaptation is performed by a reception code. Generate an adaptive excitation vector from the excitation codebook, decode the gain of the adaptive excitation vector, and obtain from the adaptive excitation vector and its gain. That adaptive power information from the change the gain quantization table excitation vector of the sound source signals, said second generating an excitation vector from the driving excitation vector codebook and decodes the gain of the excitation vector,
A speech coding / decoding method for decoding a sound source vector from the adaptive sound source vector, a driving sound source vector, and respective gains.