JPH04240897A - Speech recognizer - Google Patents

Abstract

PURPOSE:To improve the recognition rate by compensating the defects of a directional microphone and an ultra-directional microphone. CONSTITUTION:This voice recognition device is provided with the directional microphone 1, the ultra-directional microphone 12, speech recognition parts 51 and 52, registration memories 61 and 62, a CPU 8 which controls them, etc. The speech recognition parts compare voice data obtained from the microphones with voice data previously registered in the memories and obtain the code and similarity of the voice data having the highest similarity by the comparison. The CPU 8 controls speech recognition according to the code and similarity so as to output the recognition output and then the SN ratio is improved; and the direction deviation of the microphones to a sound source is prevented and the recognition ratio becomes high.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice recognition device for use in a vehicle, and more particularly to a voice recognition device that uses a super-directional microphone to improve the recognition rate.

0002

[Prior Art] Conventional pattern-matching in-vehicle voice recognition devices use
It was necessary to set the microphone as a hands-free microphone at least 50 cm away from the mouth.

0003

[Problems to be Solved by the Invention] However, as mentioned above, if a simple directional microphone is used, the S/N ratio may not be maintained, so a reduction in the recognition rate was observed in a high-noise environment. Additionally, when using a super-directional microphone to improve S/N, the speaker's sitting position or head position may shift, resulting in poor S/N. . As a result, a decrease in recognition rate could not be avoided.

OBJECTS OF THE INVENTION An object of the present invention is to provide a speech recognition device that can compensate for the drawbacks of directional microphones and super-directional microphones and improve the recognition rate.

[0005]

[Means for Solving the Problems] In order to achieve such an object, a plurality of speech recognition devices according to the first invention of the present application are provided at different positions near a speech source, and are capable of distinguishing between speech from the speech source and noise. a microphone, at least one of which is super-directional, that captures sound information from the microphone; a sound data converter that converts the sound information captured from the microphone into audio data; and a sound data converter that converts the sound information captured from the microphone into audio data during registration. Comparing the voice data at the time of recognition of the voice captured from each of the registered voice data storage unit stored together with the code and the microphone with the voice data stored in the storage unit,
Based on this comparison, a voice recognition unit that performs a recognition operation based on the code and similarity of the voice data with the highest degree of similarity, and the operation of this voice recognition unit are controlled to output a recognition result based on the code and the degree of similarity. It is characterized by comprising a recognition control section.

Furthermore, in the above device, a second invention of the present application is such that at least one of the plurality of microphones is provided at a position closest to the audio source, and all of the microphones are super-directional microphones, The voice data at the time of registration captured by the microphone is transferred to each of the registered voice data storage sections, and the recognition operation of each of the voice recognition sections is performed using the voice data.

[0007]

[Operation] The device of the present invention provides
It is possible to improve the recognition rate by providing a plurality of microphones, performing a recognition operation with a voice recognition section corresponding to each microphone, and outputting a recognition result based on the recognition operation.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the speech recognition apparatus according to the present invention will be explained in detail with reference to the drawings. FIG. 1 shows the system configuration of an embodiment of the speech recognition device according to the present invention, in which 11 is a directional microphone, 12 is a super-directional microphone, 2
1 and 22 are amplifiers, 31 and 32 are filter banks, and 41
, 42 is an A/D converter, 51, 52 is a recognition unit, 61
, 62 is a registration memory, 7 is a mode selection switch, 8 is a C
PU, 91-93 indicate data buses. In addition, amplifiers 21 and 22 are installed in order to increase the recognition rate of each microphone.
The gain is being adjusted.

FIG. 2 is a flowchart showing the basic operation of the CPU 8, and FIGS. 3 and 4 are flowcharts showing the operation of the CPU 8 at the time of voice data registration and voice recognition, respectively. Hereinafter, operations at the time of voice data registration and voice recognition will be explained based on FIGS. 1 to 4. (2) At the time of audio data registration, the "registration mode" is selected using the mode selection switch 7 (steps 101-102), and the registration routine is entered (step 103). The CPU 8 judges this and sends a registration operation command to the voice recognition units 51 and 52 (step 2).
01, 202), the voice recognition units 51, 52 wait for voice input from the A/D converters 41, 42. Recognition unit 5
1 and 52 have a built-in audio trigger, and only data between a certain level or higher input from the A/D converters 41 and 42 via the data bus is taken in as audio data.

Microphones 11, 12, amplifiers 21, 2
2, filter banks 31, 32, A/D converter 41
, 42 is always operating and transmits the voice uttered by the user.
The signal is input to microphones 11 and 12, amplified by amplifiers 21 and 22, and frequency analyzed by filter banks 31 and 32. Thereafter, the data is converted into digital data by A/D converters 41 and 42, and then taken into respective recognition units 51 and 52.

The voice data taken into the recognition units 51 and 52 are stored in registration memories 61 and 6 under the control of the recognition units.
2 and add a code number in the order of registration. (2) At the time of voice recognition, the "recognition mode" is selected using the mode selection switch 7 (steps 101, 104), and the recognition routine is entered (step 105).

The CPU 8 sends a recognition operation command to the recognition units 51 and 52 (steps 301 and 302), and the recognition unit 5
1 and 52 are waiting for voice input. When the user speaks a word to be recognized, the voice data input to the microphones 11 and 12 is sent to the recognition units 51 and 5 in the same manner as in ■.
2 (steps 303, 304), the recognition unit 5 performs DP matching with the voice data registered in If the value is too low, error information is also returned to the CPU 8. ■Voice recognition result output method A final recognition result is obtained from the two recognition results using the following method (step 305), and is output as an external device control signal (step 306). (1) When both are errors → error. (2) When two different code numbers are returned → the code number with higher similarity is selected. (3) When both return the same code number, it is assumed that the code number has been recognized. (4) When one side returns an error and the other side returns a certain code number, assume that the code number is recognized.

FIG. 5 shows the system configuration of another embodiment of the speech recognition device according to the present invention, in which 11-13 are super-directional microphones, 21-23 are amplifiers, 31-33 are band-pass filters, and 41-13 are super-directional microphones, 21-23 are amplifiers, 31-33 are band-pass filters, 43 is an A/D converter, 5
1 to 53 are recognition units, 61 to 63 are data memories, 70 is a data transfer bus, 7 is a mode selection switch, and 8 is a CPU
, 91-93 are data buses.

FIGS. 6(a) and 6(b) show an example of the arrangement of the superdirectional microphones 11 to 13, and the user has previously determined that the middle microphone 12 of the three microphones 11 to 13 is Keep it facing your mouth. The basic operation of the CPU 8 is the same as that in FIG. 2, so a description thereof will be omitted.

FIGS. 7 and 8 show details of the registration and recognition operations in the CPU 8, respectively. The operation of the embodiment shown in FIG. 5 will be described in detail below with reference to FIGS. 7 and 8. ■When registering audio data, use the mode selection switch 7 to select "registration mode". The CPU 7 determines this and sends a registration operation command to the voice recognition unit 52 (step 401).
Then, only the data portion of a certain level or higher inputted from the A/D converter 42 is taken in as audio data.

The microphone 12, the amplifier 22, the filter bank 32, and the A/D converter 42 are in constant operation, and the voice uttered by the user is input to the microphone 12, amplified by the amplifier 22, and frequency converted by the bandpass filter 32. The data is analyzed and further converted into digital data by the A/D converter 42.

The recognition unit 52 recognizes the audio data taken in via the data bus 92, stores it in the registration memory 62 via the data bus 92, and adds a code number to the data in the order of registration. It is registered in the memory 62.

Each recognition unit that has obtained the status information indicating the end of the recognition operation from the recognition unit 52 sends a control signal to each data memory 61 to 63, and transfers the data in the data memory 62 as it is to the other data memories 61 and 63. bus 7
Transfer via 0. (2) Select "recognition mode" using the voice recognition mode selection switch 7. The CPU 8 sends a recognition operation command to each recognition unit 51 to 53 (steps 501 to 503), and each recognition unit waits for input. The voice data input to each of the microphones 11 to 13 is taken into the recognition units 51 to 53 in the same manner as in (■), and the recognition units 51 to 53 perform DP matching with the voice data registered in (■), and the registered data and The code number of the registered data with the highest degree of similarity and its degree of similarity (and error information if the degree of similarity is too low) are sent to the CPU 8.
Return to. ■Voice recognition result output method The CPU 8 determines the final recognition result based on the recognition results from each recognition section using the method described below (step 507), and outputs an external device control signal (step 508). (1) When everything is an error → Error. (2) When two or more of the three return different code numbers → select the code number with the highest degree of similarity. (3) When all the same code numbers are returned → That code number is recognized. (4)1
When one or two return an error and the remaining return the same code number, the code number is recognized.

FIG. 9 shows the system configuration of still another embodiment of the speech recognition device of the present invention.
Converters 41-43, recognition units 51-53, and registration memories 61-63 are connected via data buses 91-93, respectively. The difference from FIG. 5 is that there is no transfer line from the registration memory 62. The basic operation and recognition operation of the CPU 8 in FIG. 9 are the same as the flowcharts in FIGS. 2 and 3.

FIG. 10 is a flowchart showing the registration operation of the CPU 8. The operation of the embodiment shown in FIG. 9 will be described below with reference to FIG. 10. ■When registering audio data, use the mode selection switch 7 to select "registration mode". The CPU 8 determines this and uses the voice recognition units 51 to 53.
Send a registration operation command to (steps 601 to 603)
), in the recognition units 51 to 53, the A/D converters 41 to 4
Wait for audio input from 3.

[0021] The recognition units 51 to 53 have built-in audio triggers, and only data portions of a certain level or higher inputted from the A/D converters 41 to 43 via the data buses 91 to 93 are used as audio data. take in.

The microphones 11 to 13 to the A/D converters 41 to 43 are always in operation, and the voice uttered by the user is inputted to the microphones 11 to 13, amplified by the amplifiers 21 to 23, and transmitted to the filter banks 31 to 43. 33 performs frequency analysis, and then A/D converters 41 to 43 convert it into digital data.

The audio data taken into the recognition units 51-53 is transferred to the registration memories 61-6 via data buses 91-3.
3, and at that time, a code number is added in the order of registration. (2) Select "recognition mode" using the voice recognition mode selection switch 7. The CPU 8 controls each recognition unit 5 as in the above-mentioned embodiment.
1 to 53, each recognition unit 51 to 5
3 waits for voice input. When the user utters the recognition target word, the voice data input to the microphones 11 to 13 is input into the recognition units 51 to 53 in the same manner as in ■, and the recognition unit performs DP matching with the voice data registered in ■. 51 to 53, the code number of the registered data with the highest degree of similarity to the registered data and its degree of similarity (and error information if the degree of similarity is too low) are sent to the CPU.
Return to 8. ■Voice recognition result output method The CPU 8 uses the following method to obtain the final recognition result based on the recognition results from each recognition section, and outputs an external device control signal. (1) When everything is an error → Error. (2) When two or more of the three return different code numbers → select the code number with the highest degree of similarity. (3) When all the same code numbers are returned → That code number is recognized. (4) When one or two return an error and the rest return the same code number, the code number is recognized.

[0024]

As described above, according to the present invention, in a speech recognition device, a high S/N is ensured by using a super-directional microphone, and the direction deviation with respect to the sound source can be reduced by using a plurality of microphones. Since it is prevented by using this method, a high recognition rate can be obtained.

[Brief explanation of the drawing]

FIG. 1 is a system configuration diagram of an embodiment of a speech recognition device of the present invention.

FIG. 2 is a flowchart of the basic operation of the CPU in FIG. 1;

FIG. 3 is a flowchart of registration and recognition operations of the CPU in FIG. 1;

FIG. 4 is a flowchart of registration and recognition operations of the CPU in FIG. 1;

FIG. 5 is a system configuration diagram of another embodiment of the speech recognition device of the present invention.

FIG. 6 is a diagram showing the arrangement of the microphones in FIG. 5;

FIG. 7 is a flowchart of registration and recognition operations of the CPU in FIG. 5;

FIG. 8 is a flowchart of registration and recognition operations of the CPU in FIG. 5;

FIG. 9 is a system configuration diagram of another embodiment of the speech recognition device of the present invention.

FIG. 10 is a flowchart of the registration operation of the CPU in FIG. 9;

[Explanation of symbols]

11-13 Super directional microphone 51-53 Voice recognition section 61-63 Registration memory 7 Mode selection switch 8 CPU

Claims (2)

[Claims] 1. A plurality of microphones, at least one of which is super-directional, is provided at different positions near a sound source and captures sound information of sound and noise from the sound source, and sound information captured from the microphone. a sound data converting unit that converts the sound into audio data; a registered audio data storage unit that stores the audio data at the time of registration of the audio captured from the microphones together with a code; The audio data is compared with the audio data stored in the storage unit, and by this comparison,
a voice recognition unit that performs a recognition operation based on the code and similarity of voice data having the highest degree of similarity; and a recognition control unit that controls the operation of the voice recognition unit and outputs a recognition result based on the code and the degree of similarity. A speech recognition device comprising: 2. At least one of the plurality of microphones is provided at a position closest to a sound source, and
All of the microphones are super-directional microphones, and the voice data at the time of registration captured by the microphone at the position is transferred to each of the registered voice data storage sections, and the recognition operation of each of the voice recognition sections is performed using the voice data. The speech recognition device according to claim 1, characterized in that: