JPS60218698A - Voice recognition equipment - 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 (1) Description of the field to which the invention pertains The present invention relates to a speech recognition device having a function for reducing the rate of recognition errors.

(2) Description of the conventional technology A conventional speech recognition device that registers speech percentage data as a standard turn in advance compares the registered standard pattern with the characteristic pattern of the input speech, Speech recognition was performed by selecting the standard pattern with the highest similarity to the characteristic pattern of the input speech.

However, it is known that the characteristic patterns of human speech always have uncertainties, as they differ each time they are uttered.
Even when the same word or phrase as the registration time is voice input, the degree of similarity between its characteristic pattern and the standard pattern at the time of registration differs each time the voice input is performed.

Therefore, if multiple standard patterns with high similarity to each other are registered in such a speech recognition device, due to the above-mentioned uncertainty of human speech, it will be determined that the pattern has the highest similarity to the feature pattern of the input speech. There is a drawback that the standard pattern entered may be a word or phrase different from the word or phrase actually input.

(3) Purpose of the Invention The purpose of the present invention is to provide a speech recognition device using a standard speech pattern registration method with fewer recognition errors.

(4) Structure of the Invention According to the present invention, when a standard pattern is registered, it is compared with already registered standard patterns, and if a standard pattern with a high degree of similarity exists.

A voice recognition device capable of issuing an alarm is obtained.

Hereinafter, the present invention will be explained in more detail using the drawings.

(5) Embodiment of the invention FIG. 1 shows an embodiment of the invention, in which 1 is a microcomputer, 2 is a registration key, 3 is a display, and 4 is a vocal instruction LE.
D, 5 is an A/D converter, 6 is an audio feature slam extractor, 7 is a memory, 8 is a comparator, 9 is a warning LED, 10
is a recognition key, 11 is a numeric key, and 12 is a group key. 2 and 3 are flowcharts showing the operation of the microcomputer in FIG. 1.

The registration operation of the standard pattern shown in FIG. 1 will be explained below based on the flowchart shown in FIG. When the power of the apparatus is turned on, the microcomputer 1 sets the registration number counter n to II' in step A, and enters the state of waiting for key input in step B. Here, when the registration key 2 is pressed, the microcomputer l detects this in step B, determines that the registration of the standard pattern has started in step C, and displays the value of the registration number counter n on the display 3 in step D. is displayed, the voice command LED 4 is turned on in step E, and the state of waiting for voice input is entered in step F. The point in time when the output of the A/D converter 5 exceeds a certain value due to the voice input is regarded as the beginning of the voice, and the voice feature pattern extractor 6 is operated in step H.
In step I, extract the feature pattern of the input speech.

It is determined that the value of the registration number counter n is l'1'', the voice feature pattern extracted in the memory 7 in step J is registered as the first registration pattern, and step K
At step B, 1111 is added to the value of the registration number counter n, and the process returns to step B where it waits for the next key input. The second and subsequent standard pattern registration operations from Step B to Step H are the same as the first standard pattern registration operation described above, but in Steps and P1, the microcomputer 1 uses the value of the registration number counter n to , determines that the voice feature pattern is not the first one, and operates the comparator 8 in step L.

Compare the nth voice feature pattern extracted in step H with the already registered 1st to n-1st standard patterns, and judge the result in step M,
If there are many standard patterns with high similarity, the nth voice feature pattern is registered in the memory as the standard pattern. In this step M, if it is determined that a standard pattern with high similarity exists, registration of the n-th voice feature pattern is canceled, and in step N
By lighting up the ED9, a warning is given to the voice registrant, and after steps D and E, the system enters a state of waiting for voice input in step F. Then, until it is determined that there is no standard pattern with a high degree of similarity in step M, step D is performed.
E, F, G, H.

By repeating the operations I, L, M and N, registration of similar standard patterns is prevented. After all standard patterns have been registered, press the recognition key 10, the microphone 4 computer 1 line, step C is determined to start the recognition operation,
The recognition operation shown in FIG. 3 is started.

Next, the speech recognition operation shown in FIG. 1 will be explained based on the flowchart shown in FIG. In steps O, P, Q, and R, the microcomputer 1 performs steps E, P, and R in FIG.
Perform the same operation as F, G and H. Next, step S
The comparator 8 is operated in step T to compare the standard pattern registered in the menu leaf with the characteristic pattern of the input voice extracted in step T, and based on the result, the standard pattern with the highest degree of similarity is selected in step T. Then, in step U, the registration number is displayed on the display 3, and in step 0, the system waits for the next voice input.

In addition, by changing the operation of the microcomputer 1 shown in Fig. 1, the registration number of the @similar standard pattern is notified to the user, and it is possible to register either the nth voice characteristic pattern or a registered standard pattern similar to it. You can easily add a function that allows you to choose whether to start over or classify and register into any group, or a function that automatically classifies and registers so that there are no similar standard patterns in the same group. can.

Furthermore, once all standard patterns are registered unconditionally, standard patterns are compared with each other, and if there are standard patterns with high similarity, their registration numbers are displayed, and the standard patterns specified by the user are displayed. The standard patterns specified by the user can be re-registered, or the standard patterns specified by the user can be classified into the specified groups, or the standard patterns can be automatically classified and registered so that there are no similar standard patterns in the same group. It is also easily possible to display which group the user has been registered with.

As explained above, according to the present invention, registration of similar standard patterns can be prevented in advance, and the results of comparing seven standard patterns of input voice characteristic patterns can be clearly distinguished.
Misrecognition can be reduced.

Moreover, the comparator used when registering the standard pattern can be the one used during recognition by the conventional speech recognition apparatus, so it can be easily implemented.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a configuration according to an embodiment of the present invention, and FIGS. 2 and 3 are flow charts showing its operation. l...Micro combinator, 2...
Registration key, 3... Display unit, 4... Voice instruction LED, 5... Old A/D converter, 6...
...Memory for speech feature extractor, 7..., 8...
... Comparator, 9 ... Warning LED. lO... Recognition key, 11... number key, 1
2...Groups are keys

Claims (1)

[Claims] A speech recognition device that registers voice characteristic patterns as standard turns in advance has a means for comparing the standard patterns with other previously registered standard patterns when registering the standard patterns, and similar standard patterns exist. A voice recognition device characterized by having a function of giving a warning when