JPS60203997A - Speaker shift system for 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 [Field of Application of the Invention] The present invention relates to an improvement of a speaker change method in a speech recognition device, and particularly to a speaker change method suitable for a specific speaker fp3 recognition device.

[Background of the invention]

Speech recognition technology is roughly divided into two types: speaker-independent recognition and speaker-specific recognition. Of course, speaker-independent recognition is desired, but
With current technology, only numbers can be reliably recognized, which limits the languages that can be used, and the range of applications is extremely narrow. On the other hand, specific speaker recognition has reached a level of practical use technically, but it is expensive. Therefore, in order to have the same recognition device used by many people, it is necessary to create a system that can be used by dozens of people, even if they are specific speakers. This change of speaker is called the speaker change function.

Conventionally, a method of changing speakers has been to physically change the memory in the recognition device that stores the speaker's voice recognition patterns, that is, to replace it with a floppy disk containing the speaker's own voice. However, with this method,
The operator is forced to change the floppy disk every time the speaker changes, making the operation frequent and inconvenient.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide a speaker change method for a speech recognition device that allows speaker change easily and reliably.

[Summary of the invention]

The features of the present invention include a speech recognition unit, a recognition pattern memory that stores recognition patterns set for each speaker in order to identify multiple types of speech by multiple speakers, and a machine corresponding to each speaker. comprising an operating means for generating an output signal, and a speaker identification means for inputting this machine output signal and identifying a speaker corresponding to this input, and switching the pattern for the corresponding speaker based on the speaker identification result. It is.

This makes it possible to change speakers extremely easily.

In the most desirable embodiment, the output of the speech synthesis means is used as the mechanical output signal, and the speech recognition section is shared as the speaker identification means.

This makes it possible to reliably change the speaker without requiring any special additional equipment.

[Embodiments of the invention]

FIG. 1 shows an embodiment using the present invention.

The voice input/output fixed station 1 and the voice-powered mobile station 2 are connected wirelessly.

The voice input/output fixed station 1 includes an antenna 3a, a radio 4a, a voice recognition device 5, a voice synthesizer 10, a processing device 7, and a voice recognition pattern external memory 8. The voice input mobile station 2 includes an antenna 3b, a wireless machine 4b, speech synthesis device 11,
Speaker change console 12, voicephone 14, microphone 1
Consists of 3.

In FIG. 1, there are five speakers, A to E. Information such as each person's identification number, code, name, etc. is transmitted to the voice input mobile station 2.
It is stored in the memory of the speech synthesizer 11 within the computer.

The recognition pattern for recognizing the synthesized output is the speaker recognition pattern memo I in the voice recognition device 5 in the voice input/output fixed station 1.
The voice recognition pattern is retrieved from the external memory 8 and resides in J5a. When changing speakers, the operator sets the rotary switch 15 on the speaker change console 12 to his/her name and presses the change interrupt button 16.

Due to this interruption, the speech synthesis device 11 synthesizes and outputs information corresponding to the speaker, such as the name. This synthesized mechanical output signal, for example, the voice synthesis output is the radio device 4b,
Antenna 3 of audio input/output fixed station 1 via antenna 3b
a, it reaches the radio device 4a and is received by the speech recognition device 5. The speech recognition device 5 performs pattern matching to determine which pattern in the recognition pattern memory 6a within the device matches the pattern of the received synthesized speech (mechanical output signal), recognizes the speaker, and recognizes the speaker from now on. For example, if the speaker is C, patterns of C1 to C5 are set from the voice recognition pattern external memory 8 to the recognition pattern memory 6b. Then, the speech synthesizer 10 extracts the recognized person's name from the speech synthesis data 9 and sends it to the speaker via the wireless 4a.
, is sent back to the voice input mobile station 2 via the antenna 3a.

The synthesizer receives the signal from the antenna 3b of the voice input mobile station 2,
The message is transmitted to the speaker's headphones 14 via the radio 4b, and the speaker confirms that the speaker change has been reliably performed. After that, various voice inputs are performed through the microphone 12.

Thereafter, the real voice recognition pattern to be verified by the voice recognition unit 5 is reliably switched to that of the corresponding speaker, so that sufficient voice recognition is possible even with the current level of voice recognition technology.

The operation for changing the speaker is not limited to one that requires a specific operation for this purpose, but can also be replaced by a normal operation when use is finished or a normal operation when use is started. If this is done, the speaker will always be switched to the person starting the use at the beginning of use.

The term "mechanical output signal" is used to distinguish it from an output signal corresponding to a real voice, and includes, in addition to the speech synthesis output in the above embodiment,
It may be a frequency signal in which a different frequency is assigned to each speaker, and in this case as well, the voice recognition section can be shared for speaker identification.

Furthermore, it goes without saying that various other arousal signals can be used.

〔Effect of the invention〕

According to the present invention, it is possible to easily and reliably change the speaker, and it is possible to reduce the possibility that the speech recognition device will be out of operation due to the inability to change the speaker.

[Brief explanation of the drawing]

The figure is a block diagram of a speaker change system according to an embodiment of the present invention.

Claims (1)

[Claims] 1. A means for inputting a signal corresponding to the real voice of a speaker to a speech recognition section, a speech recognition section that recognizes the input signal by matching it with a pattern, and a means for inputting a signal corresponding to a speaker's real voice to a speech recognition section; a human voice recognition pattern memory unit that stores recognition patterns set for each speaker in order to identify the voice of the speaker; an operating means that generates a mechanical output signal corresponding to each speaker; speaker identification means for identifying a speaker corresponding to the speaker, and switching a pattern of the speech recognition unit to a recognition pattern of the corresponding speaker in response to the speaker identification. Alternating method. 2. The speaker change system of the speech recognition device according to item 1, wherein the speaker identification means is the speech recognition section that compares the mechanical output signal with a speaker identification pattern set for each speaker. 3. The speaker change system of the speech recognition device according to item 2, wherein the mechanical output signal is an output of a speech synthesis means that generates a different speech signal for each speaker. 4. The speaker change system of the speech recognition device according to item 2, wherein the mechanical output signal is a frequency signal to which a different frequency is assigned to each speaker.