JPH01996A - Speech recognition method using vector quantization - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Layer] The present invention relates to a speech recognition method using vector quantization, and particularly to a speech recognition method using template matching using vector quantization.

[Prior Art and Problems to be Solved by the Invention] Automatic translation telephones use speech as input, but it is necessary to recognize the input speech. As voice recognition,
Conventionally, methods using vector quantization have been considered. In conventional speech recognition using vector quantization, the spectral distortion measure used in vector quantization has been improved in order to improve recognition performance by suppressing the increase in computational complexity and memory. A spectral distortion measure has been proposed. The significance of this method is to mix various types of features in the spectral distortion scale, use the dependencies between them as constraints, and map the features to a space with better recognition performance. However, this method has two major problems as described below.

■ In order for the dependencies between each feature to have statistical validity within the vector quantization codebook, a large number of learning samples and an enormous amount of calculation time are required.

■ In terms of codebook size, the codebook size required for each feature can be reduced by using the dependencies between features as a constraint. However, the overall codebook size is still the product of the codebook sizes required for each feature, resulting in a very large size and requiring a huge amount of memory.

Therefore, the main purpose of the present invention is to solve the conventional problems by using rate vector quantization, which generates a codebook separately for each feature as vector quantization, and performs separate vector quantization. An object of the present invention is to provide a speech recognition method using vector quantization that can solve the problem.

[Means for Solving the Problems] The present invention is a speech recognition method that performs recognition by vector quantizing input speech and comparing it with a standard pattern stored as a code string resulting from vector quantization. It has a plurality of codebooks depending on the type of voice feature, performs vector quantization on each codebook, and performs recognition using the plurality of code strings obtained.

[Operation] In the speech recognition method using vector quantization according to the present invention, by performing vector quantization on a codebook according to the type of feature of input speech, the codebook size can be changed to the codebook size required for each feature. becomes the sum of
The overall codebook size can be reduced.

[Embodiments of the Invention] Next, embodiments of the invention will be described in more detail with reference to the drawings.

FIG. 1 is a schematic block diagram of a speech recognition device according to the present invention.

In FIG. 1, the speech recognition device is composed of an amplifier 1, a low-pass filter 2, an A/D converter 3, and a processing device 4. The amplifier 1 is for amplifying an input audio signal, and the low-pass filter 2 is for removing aliasing noise from the amplified audio signal. The A/D converter 3 converts the audio signal into a 16-bit digital signal using a 12kHz sampling signal. The processing device 4 includes a computer 5, a magnetic disk 6, a terminal 7, and a printer 8. The computer 5 performs voice recognition based on the voice digital signal input from the A/D converter 3.

FIG. 2 is a flow diagram showing the overall flow from inputting an audio signal to outputting a recognition result in an embodiment of the present invention, and FIG. 3 is a flow diagram illustrating the operation of separate vector quantization. FIG. 4 is a flowchart for explaining the matching operation.

Next, the operation of one embodiment of the present invention will be described with reference to FIGS. 1 to 4. The input audio signal is amplified by the amplifier 1, and after aliasing noise is removed by the low-pass filter 2, the input audio signal is input to the A/D converter 3 in step SPl shown in FIG. 2 (abbreviated as SP in the figure). Converts the audio signal into a 16-bit digital signal.

In step SP2, the computer 5 of the processing device 4 extracts features of the audio converted into a digital signal. This feature extraction is performed using a technique such as linear predictive analysis (LPG analysis).

In step SP3, the extracted speech features are rate vector quantized by seven times with reference to the separate codebook. Separate code book is step SP
4, a codebook is prepared in advance for personal use in the case of specific speaker recognition, and as a common codebook for all speakers in the case of non-specific speaker recognition.

The code string for each feature generated by separate vector quantization has already been processed in step SP5.
At step SP6, the code string is compared with a standard pattern stored on the magnetic disk 6 as a code string, and matching is performed using the Double 5plit method, and the matching distance is sent to the result determining section 7 at step SP7. In addition,
In the case of speaker-independent recognition, the standard pattern in Ste-Tsub SP6 is created by the user uttering the recognized word in advance, and in the case of speaker-independent recognition, it is created by using a database of voices uttered by many speakers. After analysis, a representative pattern is created as a multi-template and stored on the magnetic disk 6. The result determination unit in step SP7 determines whether or not the result is appropriate for the recognition candidate, and outputs the recognition result.

Next, with reference to FIG. 3, the operations of feature extraction and separate vector quantization shown in FIG. 2 will be described in more detail. In feature extraction, in step 5P11, the audio signal converted to a 16-bit digital signal is subjected to LPG analysis using 14th order autocorrelation analysis, and the power and autocorrelation coefficient, which are the characteristics of human speech, are extracted.
Extract LPC cepstral coefficients. Step 5P12
In step 5, it is determined whether the power codebook generation is to be performed, and if it is the power codebook generation, step 5 is performed.
In P13, the power of the input voice is scalar quantized. In scalar quantization, we use the method of unconstrained m-concentration,
A power code is generated, and the generated power code book is stored on the magnetic disk 6 in step 5P14.

When a power codebook is not being generated, that is, during quantization, the power codebook in step 5P14 is used, quantization is performed in 5P15, and a code string related to power is output.

On the other hand, if it is determined in step 5P16 that the codebook is to be generated for LPG correlation coefficients and LPC cepstrum coefficients, in step 5P17, LBG
The algorithm generates a codebook based on the WLR measure, and the generated codebook is stored on the magnetic disk 6 in step 5P18. Here, L
For the BG algorithm, Linde, Buzo.

Gray:”An algorithm forV
ector Quantization Desi
gn” IEEE C0M-28 (1980-01
)It is described in. In addition, the WLR scale is a scale that emphasizes the characteristics of speech, and shows high performance in word speech recognition. (A) J64-A5 (1981-05).

Note that when the codebook of LPC correlation coefficients and LPC cepstrum coefficients is not generated, that is, when quantizing, the spectral codebook in step 5P18 is used for the autocorrelation coefficients and LPC cepstrum coefficients of the input speech, and the vector is Performs quantization and outputs a code string related to spectral information.

Here, the spectral distortion measure used for codebook generation and quantization is as follows.

d −P/P' + P' /P-2...(1
)ower d -Σ (C(n)-C' (n)) (R(
n)-R' (n))spectrum...(2) d: Distortion measure of power term over d Varnish vector distortion measure spectrum R(n): n-th autocorrelation coefficient R'(n) of codebook :Nth order autocorrelation coefficient C(n) of human power
: n-th LPC cepstrum coefficient of codebook C'(n) : n-th LPC cepstrum coefficient of input As mentioned above, when creating a standard pattern, this code string is stored as a standard pattern, and during recognition, the code of the standard pattern is Performs matching with columns.

Next, the matching method will be explained with reference to FIG. In step 5P21, Double
Matching is performed using the 5-plit method. The standard pattern in step 5P22 stores a standard pattern of power and spectrum coded by separate vector quantization. Then, in the matching in step 5P21, the distance between the codes is determined by creating a distance matrix in advance in step 5P23, and performing this table search.

In this way, the distance between the input voice and the standard pattern obtained by sequentially matching with the standard pattern is determined in step 5P.
It is output at 24.

Here, the above-mentioned matching method will be explained. In conventional matching, one feature string or code string of the input standard pattern is used, but in separate vector quantization, it is generally composed of a plurality of code strings. This embodiment uses a two-series matching method: a power code string and a spectral code string. As a distance measure considering both power and spectral information, PW
There is an LR scale. This is expressed by the following equation (3).

dPWLI? −Σ(C(n)−C′(n))(R(n
)-R' (n)) + a・(P/P' + P'/
P-2) -(3)asan0,01 In matching code strings using the conventional Double 5plit method, as mentioned above, all spaces are vector quantized and represented by a finite number of points. Then, the distances between all representative points are determined in advance and stored in a distance matrix. Therefore, d,wLR(t,j) −Dt(A(t),B(j))
DL (A (1), B (j)) one Σ (CK (n) - CL (n)) (RK (n) - RL
(n))+ a-(PK/PL+ PL/PK-2)A
(j) is the code number B of the i-th frame of the input audio. (j) is the code number DL of the j-th frame of the standard pattern. where L is the code number of A (j) and B (j).However, in separate vector quantization, there are two sequences, so the following distance is found.

d[p][νLR] (1, j) −DL (A (1), B (ms
peat 5pect 5pect""D
Lpower (Apower (1)”power
(j)) Here, DL (A (1), B (j
)) Sp(let 5pcct 5pect
−Σ(CK (n)−CL (n))(Rx (n)−
RL (n DL (A (1), B
(j)) power power
power-P/P+PL,
/P, -2 to 'L', L is the code of A (1), B (j) 5pect 5pcct number K /, Bi is the po of A (1), B (j)
is the power code number. This is obtained by separately encoding the first term and the second term of the PWLR scale, calculating the distance, and finding the sum.

Using this local distance measure, DP (dynamic
Distance is determined by matching (dynamic programming). This makes it possible to realize speech recognition using vector quantization with very high performance.

[Effects of the Invention] As described above, according to the present invention, a plurality of codebooks are provided according to the types of characteristics of input speech, vector quantization is performed for each codebook, and a plurality of obtained code strings are Since recognition is performed using , the dependence term of each feature can be ignored, the number of learning samples can be reduced, and the amount of calculation can be reduced. However, by separating, a separate vector quantization system is constructed, which slightly increases the amount of calculation, but since the number of learning samples is small, the amount of calculation can be sufficiently reduced. Furthermore, in Severt vector quantization, the codebook size is the sum of the codebook sizes required for each feature, so the overall codebook size can be drastically reduced.

[Brief explanation of the drawing]

FIG. 1 is a schematic block diagram of a speech recognition device according to the present invention. FIG. 2 is a flowchart showing the overall flow of speech recognition using the Double 5plit method. FIG. 3 is a flow diagram for explaining the operation of separate vector quantization. FIG. 4 is a flow diagram for explaining the matching operation. In the figure, 1 is an amplifier, 2 is a low-pass filter, 3 is an A/D converter, 4 is a processing device, 5 is a computer, 6 is a magnetic disk, 7 is a terminal, and 8 is a printer. -JI Im*°**−−−−−*+−−hm*−−“
°°°°°°°゛°゛°°°°゛°°°゛°°°Second
Figure 3 Input voice Figure 4 Spectrum Power code string Code T/l+ Recognition result procedure correction August 13, 1986

Claims (1)

[Claims] In a speech recognition method that performs recognition by vector quantizing input speech and comparing it with a standard pattern stored as a vector quantized code string, depending on the type of characteristics of the input speech, A speech recognition method using vector quantization, characterized in that it has a plurality of codebooks, performs vector quantization on each codebook, and performs recognition using a plurality of obtained code strings.