JPS62276598A - Speaker checking system - 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] Detailed description of the invention (Industrial application field) The present invention relates to improvements in speaker verification systems.

(Prior art) A speaker verification system detects a secret word input as voice,
This system determines whether the utterer is the person who registered the password by comparing it with the voice button of the password registered in advance and checking the degree of coincidence. Such a system is useful as a replacement for a seal in banking or as a replacement for a lock for entry security, and its realization is strongly desired.

The speaker confirmation system that has been tried in the past is described, for example, in the Journal of the Acoustical Society of Japan, Vol. 35, No. 2 (February 1978), page 63.
On page 9, each person's bus! '7-Shistw? tifl-1'l/)tcb+-shi11children5゛4ノjr,,6h,,=Le--tkoIcrshi:ly Use. In this system, during the verification operation, a code (hereinafter simply referred to as a registration number) corresponding to the name claimed by the person to be verified and a password are uttered. Next, the standard button corresponding to the inputted registration number is compared with the inputted password voice (hereinafter referred to as inputted button), and an evaluation value of the degree of difference between the distance J patterns is calculated. When this distance is smaller than a predetermined threshold, it is determined that the person to be verified is the real person, and conversely, when it is larger than the threshold, it is determined that this person is an impostor.

A problem with such a speaker verification system is that it is difficult to appropriately set a threshold value. That is, even when the same person utters a fixed password, the voice pattern differs from time to time, and the degree of variation differs from person to person. For this reason, if the threshold is set too strict (that is, to a small value), the person in question will often be determined to be the impostor, whereas if the threshold is set too loosely, the person in charge of the imposter will be mistakenly judged to be the person in question. Become.

A method of speaker verification with few such drawbacks was proposed in a patent application filed in 1983.
It is described in the specification of No. 54132. The main points of the method are as follows. That is, in addition to the standard password button (hereinafter referred to as "confirmation standard pattern"), another person's voice pattern similar to the confirmation standard pattern is registered as a suppression standard button, and the result of matching with the input button is obtained. The distance is as follows (a).

Only when the condition (b) is satisfied, the identity of the person is determined and a match signal is issued.

(B) Distance from the confirmation standard slam - Distance from the suppression standard slam (B) Distance from the confirmation standard slam - Predetermined threshold According to this method, since there is the judgment criterion (B), the threshold used in (B) is It can be set loosely. This reduces the number of erroneous determinations in which the person in question is the impostor, and also has the effect of preventing erroneous determinations in which another person is identified as the person in question (a).

(Problems to be Solved by the Invention) In the conventional method, a pattern of another person similar to a password confirmation standard pattern is selected as a suppression standard pattern. In this case, it is necessary to prepare patterns of many speakers as patterns of other people, and it is difficult to prepare speech patterns for creating suppression standard patterns when targeting the general public.

It is an object of the present invention to provide a suppression standard pattern creation method that eliminates the above-mentioned shortcomings, and to realize a speaker verification system that does not have any operational problems.

(Means for Solving the Problem) The speaker verification system according to the present invention includes a verification standard pattern storage unit that stores verification standard patterns to be registered, and a vowel pattern storage unit that stores vowel patterns of multiple speakers prepared in advance. a pressing standard pattern synthesis unit that connects the consonant part of the confirmation standard pattern and the vowel pattern to synthesize a suppression standard pattern, and stores information for synthesizing a suppression standard pattern similar to the confirmation standard pattern. a suppression designation section, an input pattern buffer that holds an input pattern uttered by a person to be matched, and a suppression standard specified by the suppression specification section based on this input input pattern and registered information input corresponding to this input pattern. The groove includes a pattern matching unit that calculates the distance to the pattern, and a determination unit that determines whether the person to be matched is the person himself/herself by using the distance to the confirmation standard pattern and the distance to the suppression standard pattern. will be accomplished.

(Operation) In the present invention, candidates for suppression standard patterns are synthesized in response to different passwords registered for each individual, and from among the candidates, a pattern that is similar to a password confirmation standard pattern is selected as a suppression standard pattern. . That is, as shown in Figure 2, the vowel sequences that appeared in the confirmation standard pattern are Vi (1), Vi (2), '', Vi (T).
- (1) Based on the fact that T is the number of vowels, replace the vowel part of the confirmation standard pattern with the vowel pattern of the majority speaker. However, the speaker of the replaced vowel pattern is different from the speaker of the confirmation standard pattern.

The vowel pattern series used for this synthesis is ■1(1)+ ■1(2)> ”°+ ■t, (T)
; Inya n-(2) (where n is the confirmed standard pattern utterance speaker). A plurality of suppression standard pattern candidates are generated. Pattern matching is performed between these suppression standard pattern candidates and the confirmation standard pattern to determine the distance. The one with the smaller distance is selected, and the pressing standard pattern candidate to which the smaller distance is given is set as the suppression standard pattern. At the time of verification, the distance between the input pattern of the person to be verified and the confirmation standard pattern and the distance between the input pattern and the suppression standard pattern are determined, and based on the two distances, it is determined whether or not the person to be verified is the person.

(Embodiment) FIG. 1 is a block diagram showing an embodiment of the present invention. The vowel pattern storage unit 130 stores vowel patterns for a large number of people in advance. For example, /aJ, /i/.

There are 10 vowel patterns such as /u/, /e/, 10/, etc.
00 people are prepared. Let be the set of these vowel patterns.

First, the operation related to the registration of the confirmation standard button will be explained. The registration information specifying section 20 consists of a keyboard, through which the information of the person to be registered, ie, the registration number n, is input, followed by the password /fujisan/. Corresponding audio is also input from the microphone 10. After this voice is analyzed by the analysis section 30,
The input button is input to the input button buffer 40 (hereinafter referred to as input button X). This manual slam X is a confirmation standard pattern An
Confirm standard button storage unit 5 in association with registration number n.
0 and stored.

Next, the composite information generation unit 100 receives the password vowel series Vi(t)(/fujisan
In the case of /, u, i, a) are taken, and the vowel sequence Vi is sent to the suppression standard pattern synthesis unit 120 as synthesis information.
(t) is sent.

Based on the vowel sequence Vnt), the suppression standard pattern synthesis unit 120 sequentially reads the vowel pattern V of speaker m from the vowel pattern storage unit 130, outputs i, and sequentially replaces the vowel part of the confirmation standard pattern. The pattern synthesized in this manner is stored in the suppression standard pattern storage section 90 as a suppression standard pattern candidate Bn.

After that, the composite information generation unit 100 changes the speaker number m from 1 to 1.
Change sequentially up to 000. However, it should be set so that the number is n. In response, the suppression standard pattern synthesis section 120 sequentially synthesizes suppression standard pattern candidates Bn, which are temporarily stored in the suppression standard pattern storage section 90 and sent to the slam matching section 60 via the signal line.

This slam matching unit 60 performs a comparison process between the input slam X and the suppression standard slam mass sent from the suppression standard slam storage unit 90, and calculates the distance D(X, Bn).
are calculated sequentially. The suppression designation determining unit 70 compares the sequentially input distances D(X, Bπ) with each other, and determines the distance from the smallest to, for example, the fifth smallest, and the corresponding number m is assigned to the suppression designation unit 80. sent to. In the suppression designation unit 80, these numbers m are associated with the registration number n as m(n, 1), m(n, 2)...m(n, 5).
...(4) The confirmation standard button An is obtained by adding f to the actual P (Oto Nana fishing), and the preparation for suppression designation is completed as shown in formula (4). This process is repeated while changing the registration number n each time a person appears.

Next, the part related to the confirmation operation will be explained. The person to be verified inputs the registration number n, which is registration information, by operating the keys of the registration number designation section 20, and also inputs /fuj from the microphone 10.
Enter the password voice: isan/. The password voice is analyzed by the analysis unit 30 in the same way as at the time of registration,
It is input to the input button buffer 40 as the input button X.

When the registration number n is inputted into the confirmation standard button storage section 50, a confirmation standard button A'' is output accordingly. Upon receiving this, the button matching section 60 first performs D(X, A).
n) distance is calculated.

When the registration number n is given to the suppression designation section 80, the suppression standard pattern designation signal m of equation (4) and the registration number n are output to the signal line k. In response to this, the suppression standard pattern storage unit 90
Suppression standard pattern B 宕; (m = m(n, 1)
.

m(n, 2), m(n, 5)) are read out.

In the slam matching section 60, the input pattern X,
A comparison is made with these suppression standard putter X width, and the distance D
(X, B';); m=m12m2. -, m5 is calculated.

The determination unit 110 compares the distance D(X, A) with the distance group of equation (4), and determines which D(X, A) is better.

Bn), and only when D(X, A) is smaller than a predetermined threshold value, it is determined that the person to be verified is the real person.

Although the principle of the present invention has been explained above based on examples, these descriptions do not limit the scope of the present invention.

In particular, in this specification, distance is used to compare voice bangs, but quantities with opposite magnitude relationships may be used, such as correlation. In this case, it is obvious that the inequality signs of criteria (a) and (b) are reversed. Furthermore, instead of (a), it is also possible to use (c) distance from the suppression standard pattern>second threshold as the determination criterion. Furthermore, in the example, the vowel sequence of the confirmation standard pattern is inputted from the outside by the keyboard by the registered speaker in order to generate synthesis information for the suppression standard pattern candidate, but all combinations are synthesized assuming that the vowel sequence is unknown. , a suppression standard pattern can also be obtained by selecting a pattern with a small distance using bang matching.

Furthermore, the suppression standard pattern synthesis unit 120 creates a suppression standard pattern by directly connecting the vowel patterns, but extrapolates the preceding and succeeding information and adds them with weights so that the vowel patterns connect smoothly in the vicinity of the connection part. It is also possible to use various speech synthesis methods such as

(Effects of the Invention) In the present invention, since the standard pattern is synthesized from the vowel patterns of many speakers, it is possible to obtain the standard pattern for any password. by this,
It is possible to realize a speaker verification system that has a higher matching rate than before and can accept any password.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIG. 2 is a diagram illustrating a method of creating a suppression standard pattern. In the figure,

Claims (1)

[Claims] A confirmation standard pattern storage unit that stores confirmation standard patterns to be registered, a vowel pattern storage unit that stores vowel patterns of a plurality of speakers prepared in advance, and a consonant portion of the confirmation standard pattern and the vowel pattern are connected to each other. A suppression standard pattern synthesis unit that synthesizes a suppression standard pattern, a suppression specification unit that stores information for synthesizing a suppression standard pattern similar to a confirmation standard pattern, and an input pattern buffer that holds an input pattern uttered by a person to be matched. and pattern matching that calculates the distance between this input pattern and the confirmation standard pattern, and the distance between this input pattern and the suppression standard pattern specified by the suppression specification section based on the registered information input corresponding to this input pattern. Department and
A speaker verification system comprising: a determination unit that determines whether or not a person to be verified is the person in question using the distance to the verification standard pattern and the distance to the suppression standard pattern.