JPS63292196A - Voice recognition equipment for specified speaker - 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 [Field of Industrial Application] The present invention relates to a speaker-specific speech recognition device, and more particularly to accurate standard pattern registration.

[Conventional technology]

Conventionally, the standard pattern registration operation in this type of speaker-specific speech recognition device has simply been to input voice from a microphone and store the obtained feature parameters in a standard pattern memory section.

[Problem that the invention seeks to solve]

However, unless you are very careful, when you make a single utterance, noise or tongue clicking may occur, and you may end up registering an unexpected standard bat. Conventional speech recognition devices for specific speakers do not have a means to confirm registered standard patterns.

By performing recognition operations using inaccurate standard patterns,
The disadvantage is that the recognition rate decreases.

[Means for resolving intermittent points]

The speech recognition device for a specific speaker of the present invention includes an A/D conversion section that converts input speech into a digital signal.

An audio memory section that stores the converted digital signal, and an audio output section that takes out the digital signal from the audio memory section and reproduces the input audio.

- Means for storing input speech in the standard pattern memory section at the time of standard pattern registration and also storing it in the voice memory section at the same time; - means for outputting human voice from the voice output section every time a word is registered or after all words are registered; have.

〔Example〕

Next, nine aspects of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing one embodiment of the present invention.

Input audio is converted into an electrical signal by microphone 1.

The audio signal is amplified by the audio input section 2. In the standard pattern registration mode, the switch 3 is connected to the contact 31, and the switch 5 is connected to the contact 51. The analysis unit 4 extracts characteristic parameters of input speech and detects speech sections. The voice section information is sent to the voice memory section 10, and the feature parameters are stored in the standard pattern memory section 6. The human voice is converted into a digital signal by the A/D conversion section 9 and stored in the voice memory section 10 according to the voice section information. The audio output section 11 takes out the digital signal stored in the audio memory section 10, converts it into an audio signal, and outputs it from the speaker 12.

In the recognition mode, the switch 6 is moved away from the attack point 31 and the switch 5 is connected to the contact 52. Therefore 2
The feature parameters extracted by the analysis section 4 are sent to the recognition section 7. The recognition section 7 compares the sent characteristic parameters with the characteristic parameters stored in the standard pattern memory, obtains a recognition result, and displays it on the second display section 8.

Next, two points of the present invention will be explained. In the standard pattern registration mode, the input audio is extracted with feature parameters,
At the same time as being stored in the standard pattern memory section 6, it is converted into a digital signal and stored in the audio memory section 10. Since the input voice stored in the voice memory section 10 is reproduced by the voice output section 11, the two speakers can easily confirm the voice registered as the standard pattern. If the registered voice is mixed with busy noise, etc., you can create an accurate standard pattern by registering it again.

The registered voice can be confirmed by controlling the voice output unit 11, and can be selected for each word or phrase of the registered voice or after all words have been registered.

〔Effect of the invention〕

As explained above, (1) the present invention has the advantage that the voice registered as a standard pattern can be easily checked and corrected at the time of standard pattern registration, so recognition is not performed using an inaccurate standard pattern, and a reduction in recognition rate is prevented. There is.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing one embodiment of the speech recognition device of the present invention. 1... Microphone, 2... Audio input section, 4... Analysis section. 6... Standard pattern memory section, 7... Recognition section, 8...
-Display section, 9...A/D exchange section, 10...audio memory section. 11...Audio output unit, 12...Speaker. Figure 1 Subi Ichiriki

Claims (1)

[Claims] 1. A voice input section that inputs voice, an analysis section that extracts feature parameters from the input voice, a standard pattern memory section that stores the extracted feature parameters, and an A/D that converts the input voice into a digital signal. a conversion unit, an audio memory unit that stores the converted digital signal, an audio output unit that extracts the digital signal from the audio memory unit and reproduces the input audio, a recognition unit that recognizes the input audio, and a standard means for storing an input voice in the standard pattern memory section at the time of pattern registration and also storing it in the voice memory section at the same time;
1. A speech recognition device for a specific speaker, comprising means for outputting input speech from the speech output section each time one word is registered or after all words are registered.