JPH04155400A - Voice recognition device - Google Patents

Abstract

PURPOSE:To prohibit the operation of either a microphone, a sound analyzer, a voice pattern generator, or a registered voice pattern memory, even in a voice registering mode under a noisy ambiance, by prohibiting the operation of either the microphone, the sound analyzer, the voice pattern generator, or the registered voice pattern memory, when the input sound level of the sound input member is larger than a specific value, in the case that the registering mode is set. CONSTITUTION:A sound input member 7 picks up the ambient noise by a voice analyzer 2 at every specific time, and when the input is finished, it is delivered to the sound input member 7 and a deciding member 8. The input member 7 decides whether there is a noise or not. The sound level from a terminal 9 is compared with a specific threshold level, and when the sound level is higher than the threshold level, a prohibition signal is generated. Since no prohibition signal is generated when there is no input of the ambient noise, a pattern of the voice analized in a pattern generator 3 is produced, and stored in a registered voice pattern memory 5. When there is the input of an ambient noise, a prohibition signal is generated, and the operation of the generator 3 is prohibited. When the voice registering mode is completed, the voice registering mode is set by a manual operation by using a mode setting member 4, and it is transferred to the voice recognition processing of a recognition processor 6.

Description

DETAILED DESCRIPTION OF THE INVENTION (A) Field of Industrial Application The present invention relates to a speech recognition device that recognizes speech in the presence of ambient sound as noise.

(b) Conventional technology Most of the speech recognition devices currently in practical use are aimed at specific speakers.In such speech recognition devices, the specific speech of a specific speaker is registered in advance. It is necessary to keep it. That is, in advance, a large number of voices are registered as patterns in the voice registration mode, and then the voice pattern input in the voice recognition mode is compared with the many voice patterns that have already been registered. Speech recognition is performed by finding the registered voice that is most similar to the input voice.

Therefore, when a speaker inputs voice in speech recognition mode, if there is any sound around the speaker, this sound becomes noise, and recognition processing must be performed on the voice buried in this noise. There was a fear that misunderstandings would occur frequently.

The adverse effects of such noise are a serious problem not only during voice input for recognition in the voice recognition mode, but also during voice input for voice registration in the voice registration mode. In other words, if you register a voice that is distorted due to noise in the voice registration mode, it will be difficult to register the voice in the subsequent voice recognition mode, even if you choose a situation without noise and carefully input the correct voice.
In the end, it becomes impossible to recognize this accurately.

(c) Problems to be Solved by the Invention The present invention has been made in view of the above-mentioned conventional drawbacks, and even in the voice registration mode in a noisy environment, the microphone, acoustic analysis section, voice pattern creation section, or An object of the present invention is to provide a speech recognition device that can prohibit the operation of any one of registered speech pattern memories.

(d) Means for Solving the Problems The speech recognition device of the present invention includes a microphone for inputting speech, a speech analysis section for analyzing speech input from the microphone,
a voice pattern creation unit that creates a voice pattern in which voice characteristics are patterned based on the analysis results of the voice analysis unit;
a registered voice pattern memory that stores, as a registered voice pattern, a voice pattern obtained from the voice pattern creation section for the voice input to the microphone in the voice registration mode;
a recognition processing unit that recognizes a voice pattern obtained from the voice pattern creation unit based on the registered voice pattern of the registered voice pattern memory for the voice input to the microphone in the recognition mode; a voice pattern memory that registers the registered voice pattern; Select a mode such as voice recognition mode for performing voice recognition based on the voice registration pattern for storing the voice in the memory. If the input sound level of the sound input section is higher than a predetermined value when setting the mode, the operation of any one of the microphone, sound analysis section, voice pattern creation section, or registered voice pattern memory is prohibited. It is something.

(E) Effect According to the voice recognition device of the present invention, when the mode setting section sets the registration mode and the input sound level of the sound input section is higher than a predetermined value, the microphone, the sound analysis section , the voice pattern creation section, or the registered voice pattern memory, thus inhibiting the creation or storage of registered voice patterns.

(F) Embodiment FIG. 1 is a block diagram showing the configuration of a speech recognition apparatus according to the present invention. In the figure, 1 is a microphone, 2 is a voice analysis section that analyzes the voice input from microphone 1, and 3 is a voice analysis section 2.
This is a pattern creation section that creates a pattern based on the analysis results.

Reference numeral 4 denotes a mode setting section for setting whether the pattern created by the pattern creation section 3 is a pattern for registration audio or a recognition audio pattern.

Reference numeral 5 indicates a registered audio pattern memory for storing the pattern created in the pattern creation unit 3 as a standard pattern when the operation mode set in the mode setting unit 4 is the registration mode; 6 indicates a memory in which the operation mode set in the mode setting unit 4 is This is a recognition processing unit that performs recognition processing of input speech when in recognition mode.

Reference numeral 7 denotes an audio input section for inputting ambient noise. For example, when the acoustic output of a car stereo becomes ambient noise, the output from the speaker is detected by a microphone and inputted to the acoustic input section 7. Alternatively, the output signal from the audio output terminal 9 of the car stereo may be input to the audio input section 7.

8 is a determination unit that determines the validity of the pattern based on the state of the audio input unit 7; in this case, it determines whether the input sound level of the audio input unit 7 is higher than a predetermined value, and based on this determination result, and controls the operation of the pattern creation section 3.

The operation of the speech recognition device of the present invention having the above configuration in the registration mode will be explained below.

Before performing voice recognition with such a voice recognition device, the user must:
It is necessary to register a plurality of voice patterns in the registered voice pattern memory 5 of the device in advance. Therefore, the user uses the mode setting section 4 to manually set the voice registration mode (switch operation, etc.).

When the registration mode is set in this way, the user speaks into the microphone 1 the voice that is to be recognized by this device. At this time, the voice analysis section 2 performs a certain period of time (for example, 5
m5ec), the audio from the microphone 1 is analyzed (for example, using an 8-channel filter bank) and the audio is extracted.

The audio input unit 7 starts capturing ambient noise in response to the start of audio extraction by the audio analysis unit 2.

The audio input unit 7 continues to capture ambient noise at fixed time intervals (for example, 5 m5ec) until the audio analysis unit 2 completes the extraction. When the voice input is completed, the voice analysis section 2 notifies the audio input section 7 and the determination section 8 of this fact. Upon receiving the notification that the voice input has ended, the determining unit 8 determines whether or not there is noise in the ambient sound input to the audio input unit 7 during the voice input period. At this time, the sound level obtained from the terminal 9 is compared with a predetermined threshold level, and when the sound level is higher than the predetermined threshold level, a prohibition signal is generated.

If the determination unit 8 determines that there is no ambient noise input to the acoustic input unit 7 during the voice input period to the extent that it adversely affects voice recognition, no inhibition signal is generated, so the voice analysis unit 2
The pattern creation unit 3 determines that the voice analyzed is valid.
A voice pattern analyzed is created and stored in a registered voice pattern memory 5.

On the other hand, if it is determined that there is ambient noise input to the acoustic input unit 7 during the voice input period to the extent that it has an adverse effect on voice recognition, a prohibition signal is generated. The operation of the creation unit 3 is prohibited.

As a result, the analysis result analyzed by the voice analysis section 2 is invalidated. At this time, it is preferable to use display means or voice response means to inform the user that the input memory of the registered voice has become invalid, and prompt the user to input the voice again.

In the above explanation, the configuration in which the operation of the voice pattern creation unit 3 is prohibited based on the determination result of the determination unit 8 was exemplified.
The storage operation of the registered voice pattern memory may be prohibited. In some cases, the microphone 1 and the acoustic analysis section 2 may
It may also prohibit any of its operations.

When the voice registration mode is completed as described above, the user manually sets the voice recognition mode (switch operation, etc.) using the mode setting unit 4, and shifts to voice recognition processing.

(G) Effects of the Invention According to the speech recognition device of the present invention, when the mode setting section sets the registration mode and the input sound level of the sound input section is higher than a predetermined value, the device Since the creation and storage operation of registered voice patterns is prohibited, it is possible to avoid registering voice patterns that are distorted due to the adverse effects of ambient noise. Therefore, since speech recognition using highly reliable registered speech patterns is possible, a decrease in the recognition rate of the speech recognition device can be avoided.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a speech recognition device according to the present invention. DESCRIPTION OF SYMBOLS 1... Microphone, 2... Voice analysis section, 3... Voice pattern creation section, 4... Mode setting section, 5... Registered voice Bakun memory, 6... Recognition processing section, 7. ...Audio input section, 8...Judgment section, 9...Terminal.

Claims (1)

[Claims] (1) A microphone for inputting voice, a voice analysis section that analyzes the voice input from the microphone, and a voice pattern creation that creates a voice pattern with voice characteristics patterned based on the analysis result of the voice analysis section. a registered voice pattern memory that stores the voice pattern obtained from the voice pattern creation section as a registered voice pattern for the voice input to the microphone in the voice registration mode; a recognition processing unit that recognizes the voice pattern obtained from the pattern creation unit as a pattern based on the registered voice pattern in the registered voice pattern memory; a voice registration mode for storing the registered voice pattern in the registered voice pattern memory; and a voice registration mode for storing the registered voice pattern in the registered voice pattern memory; A voice recognition device comprising a mode setting unit for selecting and setting a mode such as a voice recognition mode for performing voice recognition based on registered voice patterns, comprising an acoustic input unit for inputting ambient sound, and configured to set the mode. If the input sound level of the audio input unit is higher than a predetermined value when the unit is set to the registration mode, any one of the microphone, the acoustic analysis unit, the audio pattern creation unit, or the registered audio pattern memory operates. A voice recognition device that features the ability to prohibit