JP3350106B2 - Voice recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice recognition apparatus for controlling each part of a mounted apparatus by voice, and more specifically, for example, a car telephone mounted on a car, a car stereo or a navigation system, or a factory. The present invention relates to a speech recognition device used under a loud noise, such as various devices used in the above.

[0002]

2. Description of the Related Art As shown in FIG .
In, the input on the voice side and the input on the noise side are separated.
You. When this is installed in a car, the microphone for audio
A and the noise microphone a 'are usually shown in FIG.
So that it is placed before and after the driver (speaker).

[0003] A sound obtained by superposing a surrounding noise on a sound uttered by a speaker is input to a sound microphone a, and only noise is input to a noise microphone a '. The gain of the microphone amplifier b on the audio side is
It is set to a constant value and amplified by the microphone amplifier b.
The analog signal of the sound is analog / digital (A /
D) After being converted into a digital value by the conversion unit c,
In the feature extraction unit e comprising a group of
Are extracted for each frequency component, and features are extracted. Noise side
The gain of the microphone amplifier b ′ is set to a certain value.
Of the noise amplified by the microphone amplifier b '
The log signal is supplied to an analog / digital (A / D) converter c '.
After being converted to digital values in
In the feature extraction unit e ′ composed of
The feature is extracted by being separated every minute.

After that, in the voice section detection unit e, the feature extraction amount on the noise side is subtracted from the feature extraction amount on the voice side, and then the voice section is detected. During this time, the voice feature is periodically stored in the input pattern storage unit f as a digital value for each frequency component. Then, in the recognition unit g, the standard pattern stored in the dictionary unit h is compared with the feature amount of the extracted voice, and the recognition result is output from the result output unit i.

[0005]

However, in such a configuration, when the voice recognition device is used in a running automobile, the noise level depends on various factors such as road surface conditions, rain and wind, opening and closing windows, and speed. Therefore, when this noise increases, the amount of sound entering the sound microphone a is relatively reduced with respect to the amount of noise, so that detection of a sound section and sound recognition may not work well in some cases.

For example, when the engine speed is increased while the vehicle is running, the engine sound that stays inside the vehicle increases with the increase in the engine speed. Generated from room j. For this reason, the engine sound entering the microphone for sound a provided on the front side of the speaker is higher than the engine sound entering the microphone a ′ for noise provided on the rear side of the speaker.

As a result, noise reduction does not work well, and the value obtained by subtracting the noise-side feature extraction amount from the voice-side feature extraction amount may be completely different from the actual voice pattern.

The present invention has been made in view of the above-mentioned conventional problems, and has as its object to make the gain of a microphone amplifier variable according to the noise level, and to control the gain so that the sound can be reproduced. It is an object of the present invention to provide a voice recognition device capable of accurately detecting a section and performing voice recognition.

[0009]

In order to achieve the above object, an audio microphone and the audio microphone are provided.
A noise microphone placed at a different location from the
First Mai for amplifying the output of the microphones for voice
Amplifier for amplifying the output of the noise microphone.
Output from the second microphone amplifier and the first microphone amplifier.
First A / D conversion means for converting a force into a digital signal;
Converting the output of the second microphone amplifier into a digital signal;
Second A / D conversion means, and the first A / D conversion means
From the voice-side feature calculated based on the output of
Noise characteristics calculated based on the output of the A / D converter
Recognize speech by finding features of speech by subtracting signatures
Speech recognition means, comprising:
Voice microphone and noise microphone
To detect noise sources
Detecting means for detecting
The gain of the first microphone amplifier and the gain of the second microphone amplifier
Control means for adjusting and controlling the gain of the loop.

[0010]

When the voice recognition device of the present invention is mounted on, for example, an automobile, various factors (engine speed, opening of windows, etc.) considered as sources of noise (engine noise, wind noise, etc.) generated while the automobile is running. provided detection means for detecting a degree, etc.),
The first and second masks are detected based on the detection result of the detection means.
By adjusting the gain of the amplifier , voice section detection and voice recognition can be performed with high accuracy.

[0011]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows a speech recognition apparatus according to the present invention.
This speech recognition device is specifically mounted on a car and controls a car phone, a car stereo, and a navigation system. The speech microphone 1, the noise microphone 1 ', the microphone amplifier 2, 2
', A / D conversion units 3, 3', 11, feature extraction units 4, 4
, A voice section detection unit 5, an input pattern storage unit 6, a recognition unit 7, a dictionary unit 8, and a result output unit 9 as basic components, and further include a detection unit 10 and a control unit 12.

The sound microphone 1 receives a sound in which the surrounding noise is superimposed on the sound, and the sound microphone 1 'receives only the noise.

In the microphone amplifiers 2 and 2 ', the analog value of the sound converted into the electric signal by the microphones 1 and 1' for voice and noise is amplified to a predetermined level. It is controlled by information from the control unit 12 described later.

The A / D converters 3 and 3 'convert analog values input from the microphone amplifiers 2 and 2' into digital values.

The feature extraction units 4, 4 'are composed of a group of band-pass filters, and the voice-side input and the noise-side input are separated for each frequency component to extract each feature.

The voice section detection unit 5 detects a voice section after subtracting the noise-side feature extraction amount from the voice-side feature extraction amount. The input pattern storage unit 6 stores a feature amount of the voice detected in the section.

The recognizing unit 7 compares the feature amount of the voice input from the input pattern storage unit 6 with the standard pattern stored in the dictionary unit 8 in advance, and determines a standard pattern that is most similar to the voice feature amount. The result is output to the result output unit 9 as a recognition result.

The result output unit 9 sends the recognition result to each unit to be controlled by the speech recognition apparatus of the present invention, for example, a navigation system (not shown).

The detection unit 10 detects various factors such as the number of rotations of the engine and the degree of opening of windows, which are considered to be sources of noise such as engine noise and wind noise generated while the vehicle is running.
The analog value measured by the detection unit 10 is converted into digital data by an A / D conversion unit 11 and
2 is sent.

The control unit 12 calculates the noise level based on the measurement result of the detection unit 10, and adjusts the gain of the microphone amplifiers 2, 2 'based on the calculation result.

Next, the case where the number of revolutions of the engine is detected as the detection unit 10 will be described with reference to FIG.

That is, as shown in FIG.
A tachometer (rotation speed detector) for detecting the rotation speed of the engine is used, and the rotation speed of the engine is sent from the tachometer 10 to the A / D converter 11 as an analog signal. In the A / D converter 11, an analog signal from the tachometer 10 is converted into a digital value,
Sent to

The control unit 12 includes a microphone amplifier controller 13 for controlling the microphone amplifiers 2 and 2 ', and a tachometer 1
And a comparison table storage unit 14 for determining the gain setting values of the microphone amplifiers 2 and 2 ′ based on information from 0.

Thus, the information detected by the tachometer 10 is input to the microphone amplifier controller 13, where the engine speed and the microphone amplifier gain set value stored in the comparison table storage unit 14 in advance are stored. The gain setting value of the microphone amplifier is determined based on the comparison table (see FIG. 3), and the gains of the microphone amplifiers 2 and 2 ′ are controlled based on this determination.

More specifically, as shown in FIG. 3, when the rotation speed of the engine increases, the gain of the microphone amplifier 2 on the sound side is decreased, and the gain of the microphone amplifier 2 'on the noise side is increased. This creates a situation where section detection and voice recognition are easy.

As described above, by controlling the gains of the microphone amplifiers 2 and 2 'on the voice side and the noise side by feedback from the engine speed, the voice recognition rate can be increased.

Although not shown, as the detection unit 10, the following units other than the above-described tachometer can be suitably used.

[0029] That is, 1) a vibration receiver attached to the suspension of the car, 2) a lightning detector attached to the top or window or bonnet of the car, 3) a speedometer of the car, or 4) a car. Window opening / closing detector etc.
Alternatively, the detection unit 10 may include a tachometer and all of the above 1) to 4), and comprehensively determine information from these to control the gain of the microphone amplifiers 2 and 2 ′. It is also possible to configure.

When the above-mentioned items 1) to 4) are used as the detecting unit 10, the relation between these detection results and the noise level is calculated and determined in advance in the control table stored in the control table storage unit 14. What was done is used.

It should be noted that the present invention is not only for use in a vehicle for controlling a car phone, a car stereo, a navigation system and the like as shown in the figure, but also for a similar type of loud noise such as various devices used in a factory. The present invention is also applicable to control of a device used in a computer.

[0032]

As described above, according to the present invention,
The configuration is such that the gain of the microphone amplifier is controlled based on information from a detection unit that detects various factors that are sources of noise, so that voice section detection and voice recognition can be performed accurately. .

For example, when used for controlling various devices mounted on an automobile, a detection unit for detecting an engine speed, an opening state of a window, and the like, which are sources of noise such as an engine sound and a wind noise of the automobile, is provided. Based on the detection result of this detector.
Adjust the gains of the first and second microphone amplifiers
This makes it possible to accurately detect a voice section and perform voice recognition, thereby significantly improving the voice recognition rate.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a speech recognition device according to the present invention.

FIG. 2 is a block diagram illustrating a device configuration when a tachometer is used as an example of a detection unit of the voice recognition device.

FIG. 3 is a diagram showing a relation comparison table between an engine speed and a microphone up gain set value in the voice recognition device.

FIG. 4 is a block diagram showing a configuration of a conventional speech recognition device.

FIG. 5 is an explanatory diagram showing an arrangement of voice and noise microphones in an automobile.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Voice microphone 1 'Noise microphone 2, 2' Microphone amplifier 3, 3 ', 11 A / D conversion part 4, 4' Feature extraction part 5 Voice section detection part 6 Input pattern storage part 7 Recognition part 8 Dictionary part 9 Result output unit 10 Detection unit (tachometer) 12 Control unit 13 Microphone amplifier controller 14 Comparison table storage unit

Continuation of the front page (72) Inventor Izumi Kinoshita 1-3-6 Nakamagome, Ota-ku, Tokyo Inside Ricoh Company, Ltd. (56) References JP-A-3-269498 (JP, A) JP-A-57-122498 (JP, A) JP-A-58-73217 (JP, A) JP-A-3-274097 (JP, A) JP-A-58-62700 (JP, A) JP-A-3-93250 (JP, U) 62-37279 (JP, B1) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 15/20 G10L 21/02

Claims (8)

(57) [Claims] An audio microphone; and an audio microphone.
Microphone for noise placed at a different position from the microphone
Phone and amplify the output of the audio microphone
The first microphone amplifier and the output of the noise microphone
A second microphone amplifier for amplifying power, and the first microphone
A first A / D converter for converting the output of the amplifier into a digital signal
Conversion means, and a digital signal output from the second microphone amplifier.
Second A / D conversion means for converting the first A / D
Is it a voice-side feature calculated based on the output of the D conversion means?
Calculated based on the output of the second A / D conversion means.
The noise side features are subtracted to determine the voice features
Speech recognition means for recognizing
And the microphone for noise and the microphone for noise.
It is located at a different position from the crophone and is a source of noise
Detecting means for detecting the cause, and detecting results of the detecting means.
The gain of the first microphone amplifier and the second
Control means for adjusting and controlling the gain of the microphone amplifier
And a voice recognition device. 2. The control device according to claim 1 , wherein said control means detects a detection result of said detection means.
The gain of the first microphone amplifier and the second
To determine the setting value of the microphone amplifier gain and
Control table storage means for storing a reference table, and the control table storage means
On the basis of the information stored by the first microphone amplifier.
And a gain controlling the gain of the second microphone amplifier.
And an amplifier controller.
The speech recognition device according to claim 1. 3. The detection unit is mounted on a suspension unit of an automobile.
2. A speech recognition device according to claim 1, wherein said device is a vibration receiver.
Place. 4. The vehicle according to claim 1, wherein the detecting unit is located on a top of a vehicle or a window.
Or a lightning detector attached to the hood
The speech recognition device according to claim 1. 5. The vehicle according to claim 1, wherein said detecting unit is configured to rotate a vehicle engine unit.
The speech recognition device according to claim 1, wherein the speech recognition device is a number detector. 6. A method for mounting on a vehicle, wherein said detecting is performed.
2. The voice recognition device according to claim 1, wherein the unit is a speedometer of an automobile.
Place. 7. The opening / closing amount detector for a window of an automobile as the detecting unit.
Serial mounting of the speech recognition apparatus to claim 1 is. 8. The vehicle according to claim 1, wherein the detecting unit is mounted on a suspension unit of the vehicle.
Vibration receiver and the top or window of the car
Is a lightning detector attached to the hood and a car
Engine speed detector, car speedometer, automatic
It consists of an opening / closing detector for car windows.
Judging comprehensively, controlling the gain of the microphone amplifier
The voice recognition of claim 1, wherein the voice recognition is adapted to be performed.
Sense device.