JPS6091399A - 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 (al) Technical Field of the Invention The present invention relates to a speech recognition device that recognizes speech by comparing input speech with a registered speech dictionary, and particularly relates to a speech recognition device that performs speech matching every moment when necessary. be.

fbl Prior Art and Problems The speech recognition device has a registered speech dictionary that records speech parameters representing the characteristics of speech, and when speech is input, the speech recognition device calculates the speech parameters at the beginning and end speech divisions of this speech input.
The distance between the two is calculated using this voice parameter and the registered voice parameter of the registered voice dictionary, and the registered voice with the minimum distance is recognized as voice. However, in this conventional method, the input voice parameters from the beginning to the end of the input voice are determined when the voice input is completed, and the determined input voice parameters are used to check against the registered voice dictionary and convert the voice. For recognition, there was a drawback that the load on the voice processing unit increased when voice input was completed, resulting in a poor 9JJ rate.

(el) Purpose of the Invention In view of the drawbacks of the prior art, it is an object of the present invention to provide a speech recognition device that can uniformize the load on the speech processing section by performing verification every moment during the speech, and can use the speech processing section efficiently. This is the purpose.

+di Structure of the Invention Briefly stated, the present invention provides a speech recognition device that recognizes speech by comparing input speech with a registered speech dictionary, and includes a recognition start means for starting recognition when necessary, and a recognition start means for starting recognition when necessary, and a speech recognition device that recognizes speech by comparing input speech with a registered speech dictionary. A speech section control section that performs intermediate verification every moment and a selection section that selects section speech candidates obtained by the momentary matching are provided, and the speech is sequentially recognized from the section speech candidates to perform the required speech recognition. It is characterized by:

(Examples of 81 inventions Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the speech recognition apparatus of the present invention, and FIG. 2 is a related diagram showing the number of frames of speech input.

In the figure, 1 is a voice sampling section, 2 is a parameter extraction section, 3 and 4 are registers, 5 is a voice section control section, 6 is an operation panel, 7 is a reception section, 8 is a collation section, 9 is a registered speech dictionary, 10 indicate candidate selection sections, respectively.

The audio input is sampled by an audio sampling section 1, and this sampling value is input to a variation extraction section 2. The parameter extraction unit 2 extracts parameters indicating characteristics of the voice and stores them in the register 3. The stored parameters are compared with the speech parameters in the registered speech dictionary 9 in the matching section 8, and this point is a feature of the present invention.

That is, when a key to start recognition is pressed on the operation panel 6, the receiving section 7 accepts the start of recognition and notifies the voice section control section 5 to that effect. As shown in FIG. 2, the voice section control section operates to divide the time between the start and end of the voice input into required frames and collate P1, F2 to Fn. That is, at time t1, the input audio barata of frame F1 is taken out from the register 3 and inputted to the matching section 8. On the other hand, the voice section control unit 5 indicates in the registered voice dictionary 9 that the voice is an Fl frame, and sequentially inputs the registered voice of the Fl frame to the matching unit 8. Therefore, the matching unit 8 performs matching with all registered voices.

What are the verification results? ! It is input to the selection section 10 and the selection? The ili selection unit 10 sequentially selects a required number of candidates, for example, 1/2 of the number of registered voices, starting from the shortest distance, and stores these registered voice parameters in the register 4.

At time t2, the voice input parameters between tl and t2 stored in the register 3 are compared with all of the candidate registered voice parameters described above, and candidates are selected and stored in the register 4.

In other words, the F2 frame is compared, and the candidates are 1/2 x 1/
2, or 1/4. The sequential process is repeated until the required speech recognition is completed. By operating the operation panel 6, verification can be performed by dividing into required sections.

(f) Effects of the Invention As explained in detail above, the speech recognition device of the present invention is capable of equalizing the load on the speech processing section by performing verification moment by moment during the speech, resulting in efficient use of the speech processing section. Therefore, it has many advantages when performing speech recognition.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the speech recognition apparatus of the present invention, and FIG. 2 is a related diagram showing the number of frames of speech input. In the figure, 1 is a voice sampling section, 2 is a parameter extraction section, 3 and 4 are registers, 5 is a voice section control section, 6 is an operation panel, 7 is a reception section, 8 is a collation section, 9 is a registered speech dictionary, 10 indicate candidate selection sections, respectively.

Claims (1)

[Claims] A speech recognition device that recognizes speech by comparing the input speech with a registered speech dictionary, includes a recognition start means for starting rope weaving when necessary, and a voice section control section that performs comparison every moment during utterance by operating the recognition start means. What is claimed is: 1. A speech recognition device comprising: a selection unit for selecting segment speech candidates obtained through verification performed moment by moment, and sequentially recognizing speech from the segment speech candidates to perform required speech recognition.