JPS5922100A - 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 TECHNICAL FIELD The present invention relates to speech recognition under adverse conditions, and in particular to an improvement of a speech recognition device under background noise.

<Prior art> Generally, it has been suggested that the configuration conditions for a speech recognition system include individual differences, input categories, speaking speed, units of pronunciation, and conditions for reducing speech and its background noise (19
Published March 10, 1972, Speech Science, pp. 370-3
72 pages, Publisher: University of Tokyo Press).

It is also stated that an experiment was conducted in which audio-based automatic sorting of parcels carried on a conveyor under high noise levels was conducted. These experimental facts suggest that limited word recognition of a few to several dozen words has reached a practical level even under adverse conditions (December 25, 1970, Voice, No. 145) Pages 146 to 146, Publisher: Corona Publishing).

However, speech recognition under such adverse conditions is significantly affected, especially by background noise.

In this case, there is a possibility that instantaneous high noise may occur.
This may be due to the analysis of human speech, but due to the influence of considerable noise, the correct standard pattern was not created during speech registration, which was a major cause of misrecognition.

The same thing can be said about the input voice during recognition.For example, in order to determine whether the input voice is actually recognized or not, it is possible to play back the registered voice and take advantage of the trial listening effect. 11] Is it possible to completely cut off the level of speech recognition?Is there a problem with this, as it would be too complicated and would make simulation difficult?

<Object of the Invention> The present invention has been made in order to solve the above-mentioned conventional problems, and provides a speech recognition device that can be realized by reducing the negative effects of speech recognition in the background miscellaneous H', -1: We aim to provide the following.

In particular, the present invention uses a microphone for noise detection separately from the human voice microphone, and uses a display device to check whether the fIF sound level from this microphone satisfies condition 1/Q for voice recognition. In addition to indicating digits, if a background sound (external noise) that interferes with the normal operation of the device is detected during use, if voice registration is in progress, voice registration will be prohibited when noise occurs, and the user will be notified. The present invention provides a speech recognition device which has a function of instructing voice input again and, if recognition is in progress, not performing the voice recognition and instructing voice input again.

<Configuration of an Embodiment of the Invention> An embodiment of the invention will be described in detail below with reference to the drawings.

The drawing is a block diagram showing one embodiment of the present invention. In the figure, 1 is an audio input microphone, and 2 is an II installed at a location separate from the audio input microphone.
This is a microphone for detecting F sound (noise). This microphone is configured to receive only external (background) noise. The input sounds of the side microphones 1 and 2 are amplified by 1 in amplifiers 3 and 4, respectively, and the noise level of the amplified 111ti output of the noise detection microphone 2 is calculated in the power calculation unit 5 and used by the device. The control section 8 determines whether or not there is a loud noise N, based on the output result of the power calculation section 5, and displays it to the user through the display device 6, for example, an alphanumeric link type liquid crystal display device. It is possible to indicate that it is in a noise zone. Reference numeral 7 denotes an analysis circuit that analyzes the voice input from the voice input microphone 1.The standard pattern resulting from the analysis is registered in the registration memory 10 if it has a noise level that can be used, and if it cannot be ignored at the time of voice recognition, it is not ignored. If noise is detected, the control unit 8 is controlled based on the calculation result of the power calculation unit 5, the processing in the recognition unit 9 is performed, and the user is instructed to input 1] through the display device 6. give. In this way, one-shot recognition caused by the influence of disturbances due to noise can be avoided. What about the registration memory 10? 1' conductor memory, for example, a W6 output [hard random access memory (RAL [
) can also be used, and the control section 8 can be a microprocessor.
It can be realized in a formal manner using a microcomputer or a microcomputer.

It goes without saying that the present invention can be applied to, for example, electronic equipment such as computer processors, computer machines, and electronic meters.

(Effects of the invention) As stated above, the present invention can effectively realize voice recognition in background Ml (accounting), especially voice registration,
It has the advantage that it is possible to visually determine whether or not the device is in use at the time of recognition.

[Brief explanation of drawings]

The drawings show l'r'I"' ll according to one embodiment of the present invention.
, represents a block diagram of the S recognition device. 1: Microphone for audio input, 2: Microphone for 4" It g detection, 5: Power calculation section, 7: Input circuit, 8
: Control unit, 9: Recognition unit, 10: Registration memory.

Claims (1)

[Claims] (1) The audio input microphone has a noise detection microphone provided individually, and a noise level detection means for detecting the noise level of the noise detection microphone, and the noise level detection means A voice recognition device characterized in that it has a function of instructing re-registration if voice is being registered, prohibiting recognition during voice recognition, and instructing voice input again based on the detection level of the means.