JPH09127982A - Voice recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To eliminate the complexity to confirm two microphones for voice input and noise input before usage by using one of two microphones for voice input when the other is used for noise input, and using the other for voice input when one is used for noise input. SOLUTION: When voice is generated toward a microphone 1, a voice interval detection section 3 detects the end of the voice interval, and the feature quantity outputted from a feature extraction section 5 is selected by a selection section 7. When voice is generated toward a microphone 2, the end of the voice interval is detected by a voice interval detection section 4, and the feature quantity outputted from a feature extraction section 6 is selected by the selection section 7. A recognition process section 8 can separately receive the feature quantities in individual voice intervals. Either one of two microphones can be used for voice input while two microphones are not distinguished for voice input and noise input.

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, and more particularly, to a voice recognition device that performs two-input voice detection by two microphones to cut out a voice section and recognize a voiced voice.

[0002]

2. Description of the Related Art As a first conventional technique for recognizing uttered voice by two-input voice detection using two microphones, there is a "voice section detecting method" disclosed in Japanese Patent Laid-Open No. 62-42197. The "speech segment detection method" which is the first conventional technique will be described with reference to the block diagram of FIG.
A main microphone 101 installed so that the N ratio is large, and a sub microphone 102 installed so that the SN ratio is smaller than that of the main microphone 101.
A main microphone pre-processing unit 103 that pre-processes an input signal from the main microphone 101; a sub-microphone pre-processing unit 104 that pre-processes an input signal from the sub microphone 102; and a power of an input signal from the main microphone 101. And a main microphone power calculator that
A sub microphone power calculation unit 106 that calculates the input power from the sub microphone power calculation unit 104 that calculates the power of the input signal from the sub microphone 104, and a noise level power calculated by the main microphone power calculation unit 105 during noise learning. threshold P _T and the noise learning unit 107 of setting, the threshold time for a predetermined time and set threshold P _T by noise learning unit 107 threshold for speech interval calculated based on the t _T
Voice section candidate detection unit 1 for detecting voice section candidates based on
08, a level difference (or ratio) detection unit 109 that obtains a difference (or ratio) between the output of the main microphone power calculation unit 105 and the output of the sub microphone power calculation unit 106, and a voice based on the output of the level difference detection unit 109. Section candidate detection unit 108
A voice section determination unit 110 that determines whether to adopt or reject the voice section candidate that is determined as described above, and a voice recognition unit that recognizes the voiced voice of the voice section determined by the voice section determination unit 110. And 111. In this first conventional example, the main microphone 101 of the two input microphones is installed 20 cm in front of the speaker, and the sub microphone 10
2 is installed at a position of 80 cm in width. Also, both microphones are installed so that ambient noise is input under similar conditions. In FIG. 3, first, the power level of the ambient noise before the voice is input is learned by the noise learning unit 107 using the main microphone 101. Thereafter, the input power level of the main microphone 101 is continuously monitored, and the level in the voice section candidate 108 is set to the threshold P set by the noise learning unit 107.
_{When the} time that is longer than _T is equal to or longer than the threshold t _T , the section is set as a voice section candidate, and when the condition is not satisfied, the above operation is continued. When a voice section candidate is found, the level difference detection unit 109 detects the average power level (db) of the input from the main microphone 101 and the average power level (db) of the input from the sub microphone 102 in the section. The difference is calculated, and the voice section determining unit 110 determines the difference by a predetermined threshold T
When it is larger than _lem , the section is set as a voice section, and when the condition is not satisfied, the voice candidate section is rejected, and the operation returns immediately after the noise level learning. Here, the power level threshold P _T is a value obtained by adding a predetermined appropriate value to the learned power level of the ambient noise. The time _threshold t _T and the power level difference _threshold T _lem are respectively predetermined values.

Next, 2 using two microphones
Japanese Laid-Open Patent Publication No. 58-196599 discloses second and third conventional techniques that can perform recognizing processing of voicing voice satisfactorily even in an environment where there is noise around the input voice.

Generally, in order to perform voice recognition in an environment where there is noise, a noise microphone 202 is provided in addition to the voice microphone 201 as shown in FIG. Of the output signal of the
After passing through, the subtractor 205 performs subtraction processing with the output signal from the voice microphone 201, and then the output of the subtractor 205 is added to the feature extraction unit 206 to extract the feature portion. It is stored in the storage unit 207 and stored in the storage unit 207. At the time of voice recognition, a signal from the feature extraction unit 206 is collated by the recognition unit 208 with the feature unit stored in the storage unit 207 by a pattern matching method. ing. That is, in the voice recognition device according to the second conventional technique, the speaker 209 speaks a word to be recognized by the voice microphone 201 and delays the output signal of the noise microphone 202 from the output signal of the voice microphone 201. Delay with device 203,
The gain controller 204 performs gain control and then subtracts the gain to remove sounds (noise) other than the voice uttered by the speaker 209, and perform voice recognition using the signal. However, in the second prior art speech recognition apparatus, the phase shift caused by the difference in the transmission distance from the two noise sources to the microphones 201 and 202 can be compensated by the delay device 203. The device 203 is only for a single noise source, and when there are a plurality of noise sources in different places, the phase shift cannot be corrected completely, and there is a drawback that voice recognition is difficult to perform satisfactorily. Further, when the voice recognition device is used in a narrowly closed space such as an automobile, the output signals of the voice microphone 201 and the noise microphone 202 due to noise generated from the noise source are not limited to the level difference. However, since the frequency changes considerably depending on the sound field characteristic, there is a drawback that the noise cannot be sufficiently removed only by performing the gain control by the gain controller 204, and the voice recognition is difficult to perform satisfactorily.

The third and prior art speech recognition devices are intended to eliminate the drawbacks of the second prior art, in which the output signals of the voice macrophone and the noise microphone are respectively extracted by the feature extraction unit. At the same time as extracting the characteristic portion, the gain control is performed for each channel constituting the characteristic extracting portion with respect to the output signal of the noise microphone, and the difference between the data extracted by the characteristic extracting portion is obtained so that the voice recognition is performed by pattern matching. It is composed. According to such a configuration, not only the level difference between the output signals of the voice microphone and the noise microphone, but also the frequency characteristic caused by the sound field characteristic can be recognized without being affected by the background noise. This has the advantage that good speech recognition can be performed in a noisy environment. That is, the phase information of the signal has no significant meaning in speech recognition, and the phase information is discarded in the feature extraction unit. Therefore, even if the features of the output signals of the voice microphone and the noise microphone are extracted by the feature extraction unit and then the difference between the two signals is obtained and pattern matching is performed with the data, the voice is not affected much by the background noise. Can recognize. Further, since it is possible to absorb the level difference and the difference in frequency characteristic by individually controlling the gain of each channel signal of the feature extraction unit for the output signal of the noise microphone, noise can be removed more strictly, Stable voice recognition can be performed even in a place where background noise is complicated.

Referring to the block diagram of FIG. 5,
The output signals of the voice microphone 310 that collects voice waves and the noise microphone 311 are separated by the feature extraction unit 3.
12 and 313 perform feature extraction, and feature extraction signals obtained by performing feature extraction from the output signal of the noise microphone 311 are subjected to gain control by the gain controller 314 for each channel so that noise can be best removed, and then the subtractor 315. In addition to the above, the subtracter 315 subtracts from the signal of the voice microphone 310 whose feature is extracted. At the time of registration, the subtractor 315 and the output are sent to the storage unit 316, and at the time of recognition, the output of the subtractor 315 is sent to the recognition unit 317, and the storage unit 3
Pattern matching with the data registered in 16 is performed, and the closest one is output as the recognition result.

[0007]

In these conventional voice recognition devices, a voice microphone and a noise microphone are distinguished from each other, and it is troublesome to check which one of the voice input microphones is used. there were.

Further, even if the voice microphones are visually distinguished, the connection to the device may be erroneously reversed, so that confirmation work is indispensable, and it is easy to make a mistake in using the device. was there.

[0009]

A voice recognition device according to the present invention is a voice recognition device for performing a two-input voice detection by a first and a second microphone to recognize and process a uttered voice. Is used for voice input, the second microphone is for noise input.

The voice recognition apparatus according to the present invention is also the first
And a voice recognition device for recognizing an uttered voice by performing two-input voice detection by a second microphone, wherein the first voice section detecting means outputs the uttered voice when the voice is generated toward the first microphone. The first microphone outputs the first voice signal as the first voice signal to the voice input end and the second microphone outputs the second voice signal.
From the output end of the voice input to the noise input end as the first noise signal, the voice section of the voice uttered toward the first microphone is detected, and the first voice signal of the voice section is cut out to obtain a third voice signal. Is output as a second voice signal from the first output end of the second microphone and the voice output when the second voice section detecting means outputs the voice signal toward the second microphone. A voice section of the voice uttered toward the second microphone is detected by inputting to the noise input end as a noise signal from the second output end of the first microphone as well as the second output end of the first microphone. The second voice signal in the voice section is cut out and output as a fourth voice signal, and is uttered toward each microphone based on the third voice signal and the fourth voice signal. And recognizes processing each respective utterance voice.

Further, the speech recognition apparatus according to the present invention is a speech recognition apparatus for recognizing an uttered speech by detecting two input speeches by the first and second microphones,
The first voice section detection means inputs the voice output when the voice is uttered toward the first microphone from the first output end of the first microphone to the first voice input end as a first and voice signal. And the voice of the voice uttered toward the first microphone, which is input from the second output end of the second microphone as the first noise signal to the second voice input end through the first delay means. A section is detected, and the first voice signal in the voice section is cut out to obtain a third section.
Voice signal when the second voice section detecting means utters the second microphone from the first output end of the second microphone. As a second noise signal from a second output end of the first microphone as a second noise signal through a second delay means to the second voice input end as the second microphone. To the third voice signal and the fourth voice signal by detecting the voice period of the voice uttered toward the user, cutting out the second voice signal in the voice period, and outputting the second voice signal as a fourth voice signal. It is characterized in that each of the uttered voices uttered toward each of the microphones is recognized based on the above.

Furthermore, the voice recognition apparatus according to the present invention includes a first microphone and a second microphone, which are installed close to each other, for converting a vocal sound into an electric signal.
The first electric signal output from the first output end of the first microphone is input to the first input end as a first audio signal and is output from the second output end of the second microphone. The second electric signal is input to the second input terminal as the first noise signal, and is directed to the first microphone only when the level of the first voice signal is higher than the level of the first noise signal. Said first voiced and output
First voice detection means for detecting a voice section of the voice signal as a first voice section and cutting out the first voice signal of the first voice section and outputting it as a second voice signal, The second electric signal output from the first output end of the second microphone is input to the first input end as a third audio signal and output from the second output end of the first microphone. The first electrical signal that is
The second noise signal is input to the input end of the third voice signal and the third voice signal is output toward the second microphone only when the level of the third voice signal is higher than the level of the second noise signal. Second by detecting the voice section of the voice signal of
Second voice detecting means for cutting out the third voice signal in the second voice period and outputting it as a fourth voice signal, and the third voice signal from the first voice detecting means. Of the fourth voice signal from the second voice detecting means, and a first feature extracting means for calculating a feature amount of the voice signal of Second feature extraction means for calculating a quantity, converting it into a feature vector series, and outputting it as a second feature vector series; the first feature vector series and the second feature vector series input from the first feature extraction means; Selecting means for comparing the second feature vector series input from the feature extracting means of 1), selecting the feature vector series of which the end of the voice section is detected first, and outputting as the third feature vector series. And before And a third feature recognition processing means from the vector sequences recognizing process an utterance voice uttered toward the first microphone or the second microphone input from the selection unit.

Furthermore, in the voice recognition device according to the present invention, the first voice detection means outputs the second electric signal output from the second output end of the second microphone to a predetermined first. A level of the first audio signal input as a fifth audio signal to the second input terminal through a first delay means having a delay time and input from a first output terminal of the first microphone; The third voice segment is extracted by comparing the voice level of the fifth voice signal with the voice segment of the voice signal with the higher level to determine the third voice segment and the third voice segment is cut out. Second delay signal having a predetermined second delay time, the first electrical signal being output as a voice signal of the first voice signal and the second voice detecting means outputting the first electrical signal from the second output end of the first microphone. The second through the means The sixth voice signal is input to the input end as a sixth voice signal, and the level of the second voice signal input from the first output end of the second microphone is compared with the level of the sixth voice signal. The voice section of the larger voice signal is detected as a fourth voice section, and the voice signal of the third voice section is cut out and output as the fourth voice signal.

[0014]

Next, the present invention will be described with reference to the drawings. Referring to FIG. 1 showing a first embodiment of the present invention, a voice recognition device converts a voiced voice input from a voice input person into an electric signal and outputs a voice signal a from an output end A and a voice signal from an output end B. A unidirectional microphone 1 that outputs the signal b at the same impedance and the same level for each, and the microphone 1 converts voiced voice input from a voice input person into an electric signal in a different voice input direction. A unidirectional microphone 2 for outputting a voice signal c from the output end A and a voice signal d from the output end B at the same impedance and the same level, and a microphone 1.
Of the voice signal a input from the output end A of the microphone 1 to the voice input end C of the voice input person to the microphone 1 and the voice signal c input from the output end B of the microphone 2 to the noise input end D. A voice section detection unit 3 that detects a voice section of a voice signal a input from the microphone 1 as a noise component and outputs the voice signal of the cut voice section as a voice signal e, and an output end B of the microphone 1. The voice signal b input to the noise input terminal D from the microphone 2 is detected as a noise component, and the microphone 2
The voice signal c input from the output end A to the voice input end C is detected as the voiced voice of the voice input person to the microphone 2, and the voice section of the voice signal c input from the microphone 2 is specified and cut out. A voice section detection unit 4 that outputs a voice signal of a voice section as a voice signal f, and a feature amount of a voice signal e input from the voice section detection unit 3 are calculated and converted into a time-series feature vector to obtain a feature vector series g. And a feature extraction unit 5 that calculates a feature amount of the voice signal f input from the voice section detection unit 4 and converts the feature amount into a time-series feature vector and outputs the feature vector sequence h. Feature vector sequence g input from the extraction unit 5 and feature vector sequence h input from the feature extraction unit 6
Are compared at the same time, and the feature amount of the one whose end of the voice detection result is detected earlier, that is, the one of the end of the voice signal e and the end of the voice signal f, which is detected earlier, is the feature amount of the recognition target. A selecting unit 8 for selecting a feature vector sequence to output as a feature vector sequence i, and a selecting unit 8
The recognition processing unit 8 performs recognition processing on the feature amount of the feature vector series i input from the recognition processing unit 8 for recognizing voiced speech input to the microphone selected by the selection unit 8.

With the above configuration, when a voice is uttered toward the microphone 1, the end of the voice section is detected first by the voice section detection unit 3, so the feature amount output from the feature extraction unit 5, that is, the feature. When the vector sequence g is selected by the selection unit 7 and a voice is uttered toward the microphone 2, the end of the voice section is detected first by the voice section detection unit 4 and is output from the feature extraction unit 6. Since the feature amount, that is, the feature vector sequence h is selected by the selection unit 7, the recognition processing unit 8 must individually accept the feature amount of each voice section regardless of which microphone the voice is emitted toward. Therefore, both microphones can be used for voice input without distinguishing the two microphones for voice input and noise input.

More specifically, both the microphone 1 and the microphone 2 are unidirectional microphones having the same output impedance. For example, when the microphone 1 is installed vertically upward, the microphone 2 is installed horizontally horizontally. On the contrary, when the microphone 2 is installed vertically upward, the microphone 1 is installed horizontally horizontally. That is, each of the two microphones is installed so that the directions of the two microphones open 90 degrees, and when the voice uttered toward one microphone is input to the other microphone, the two microphones are installed so as to have a noise equivalent level due to its directivity. . The distance between the two microphones is, for example, about 60c
m.

First, the operation for confirming the voice uttered toward the microphone 1 will be described. Now, assuming that a voice is uttered toward the microphone 1, the voice signal a output from the output end A of the microphone 1 is input to the voice input end C of the voice section detection unit 3 and the output end B of the same microphone 1 is input. The voice signal b output from the same is input to the noise input terminal D of the voice section detection unit 4 at the same level as the voice signal a. On the other hand, although the voice uttered toward the microphone 1 reaches the microphone 2, the voice generated toward the microphone 1 which is actually input to the microphone 2 is directed to the microphone 2 due to the directivity depending on the installation direction of the microphone 2. The level that is electrically converted and output as the audio signal c and the audio signal d is about noise level. The voice uttered toward the microphone 1 in this way is input from the output end A of the microphone 2 to the voice input end C of the voice section detection unit 4 as the voice signal c, and from the output end B as the voice signal d. The noise is input to the noise input terminal D of the detection unit 3. That is, since the voice signal C having the noise level input to the voice section detection unit 4 is input to the voice input terminal C, it is treated as voice,
Further, since the voice signal d having the noise level input to the voice section detection unit 3 is input to the noise input terminal D, it is treated as noise. Then, in the voice section detection unit 3, first, the level of the voice signal a as the voice level from the microphone 1 and the level of the voice signal d as the noise level from the microphone 2 are compared, that is, the voice signal a is S (signal).
Also, the S / N ratio, that is, the signal-to-noise ratio is obtained with the voice signal b as N (noise), and the S / N ratio is the voice section detection unit
When the S / N ratio threshold is smaller than the predetermined threshold internally, the sound signal a input to the sound input terminal C is regarded as a normal sound signal uttered toward the microphone 1 and, on the other hand, the S / N ratio threshold is When it is larger than the threshold, it is determined that the audio signal a is not regarded as a regular audio signal. Therefore, when a voice is generated toward the microphone 1, the level of the voice signal a input to the input end C of the voice section detection unit 3 is naturally input to the input end D of the microphone 2.
Is much higher than the level of the audio signal d from the S / N ratio, and its S / N ratio is certainly smaller than the S / N ratio rationalized. Therefore, the audio signal a can be easily generated as a normal sound generated toward the microphone 1. Can be determined.
On the other hand, in the voice section detection unit 4, the noise level voice signal c from the microphone 2 input to the voice input end C is more than the normal voiced voice level b from the microphone 1 input to the noise input end D. Since it is much smaller and its S / N ratio is certainly larger than the S / N ratio threshold, it can be easily determined that the voice signal c is not the normal voice generated toward the microphone 2. .

In the voice section detection unit 3 which determines that the voice signal a input to the voice input terminal C is a normal voice, the voice section of the voice signal a to be input next every frame period and the start end of the voice signal a and In order to obtain the termination, the noise level from the noise level input to the noise input terminal D including ambient noise when no sound is generated in advance in either the microphone 1 or the microphone 2 within a predetermined time of the noise level, that is, The noise threshold is calculated by calculating the average power of a section traced back a certain time from the current time, and the calculation of the noise threshold is repeated at regular intervals until the average power of a of the audio signal exceeds this noise threshold. , When the average power of the audio signal a exceeds the latest noise threshold calculation value, it is recognized as the beginning of the audio signal a, and the noise threshold is detected. A value obtained by multiplying the audio signal by a specific coefficient is subtracted as the noise power from the average power of the audio signal a, and once the start end is recognized, the noise threshold is fixed and kept until the end of the audio signal b is recognized. . Then, when the average power of the audio signal a becomes smaller than the noise threshold, the end of the audio signal a is recognized, and once the end is recognized, the calculation of the noise threshold is started again from that time, Thereafter, the same operation as described above is repeated to detect the voice section of the voice signal a and its start and end, and output it as the voice signal e.

On the other hand, during this period, the voice section detection unit 4 does not consider the voice signal c input to the voice input terminal C to be a normal voice, so that the voice section and its start and end are not detected, and therefore the voice section is not detected. The voice section detection unit does not output the voice signal f in the section in which the detection unit 3 outputs the voice signal e.

On the contrary, when the voice signal c input to the voice input terminal C of the voice section detection unit 4 is determined as a normal voice, that is, when the voice is uttered toward the microphone 2, the voice section detection is performed. Similarly to the above, the section 4 detects the voice section of the voice signal c and its start end and end, and outputs the voice signal f. The voice section detecting section 3 detects the voice signal a input to the voice input terminal C at this time. Since it is not regarded as a normal voice uttered toward the microphone 1, the voice section and its start end and end are not detected, and therefore, this section, that is, the voice section detection unit 4 outputs the voice signal f. The audio signal e is not output in the section.

As described above, regardless of whether the voice is uttered toward either the microphone 1 or the microphone 2, in the voice section detecting section on the microphone side where the voice is generated, the voice generated toward the microphone of the own system is generated. Accurately detect the voice section and its beginning and end,
It is possible to cut out the vocalized voice signal in that section.

Next, in the feature extraction unit 5, the voice section detection unit 3
The audio signal e cut out by
A time-axis voice waveform is converted into a frequency component by data conversion such as ET, and the converted frequency component is further converted into a time-series feature vector by calculating a spectrum envelope by logarithmic spectrum conversion or the like. A feature vector sequence g by adding start and end information of a voice section
Output as The feature extraction unit 6 also functions in the same manner, converts the voice signal f cut out and input by the voice section detection unit 4 into a time-series feature vector, and adds the start and end information of the voice section to the feature vector. Output as a vector series h. The selection unit 7 constantly monitors the feature vector sequence g input from the feature extraction unit 5 and the feature vector sequence h input from the feature extraction unit 6, and the voice added to each feature vector. The feature vector series whose end information of the section is detected earlier is selected, output as the feature vector series i, and supplied to the recognition processing unit 8.

Since the technique of recognizing the voiced speech from the feature vector sequence in the recognition processing unit 8 is a known technique, its description is omitted.

Next, a second embodiment of the present invention will be described. Referring to the block diagram of FIG. 2 which is the second embodiment, the microphone 1, the microphone 2, the feature extraction unit 5, the feature extraction unit 6, the selection unit 7 and the recognition processing unit 8 are the same as those in the block diagram of FIG. Since each block has the same function as the block of the same symbol in the embodiment, the description thereof is omitted. First
5 is different from that of the first embodiment in that the voice signal b output from the output end B of the microphone 1 is input to the voice input end D of the voice section detection unit 10 through the delay circuit 12 and output from the output end B of the microphone 2. 2 is input to the voice input terminal D of the voice section detecting unit 9 through the delay circuit 11, and each noise input terminal D in the voice section detecting units 3 and 4 in FIG. The detection units 9 and 10 have the same structure as the voice input end C as each voice input end as described above.

More specifically, the voice uttered toward the microphone 1 is input from the output end A of the microphone 1 to the voice input end C of the voice section detection unit 9 as the voice signal a and from the output end B of the microphone 2. It is output as a voice signal d having a noise level, delayed by a delay circuit 11 for a predetermined time, and input as a voice signal d ′ to a voice input terminal D of the same voice section detection unit 9. Then, the voice section detection unit 9 compares the levels of the voice signal a and the voice signal d ′, and when the level difference is larger than the internally determined level difference threshold, the higher level, That is, here, the audio signal a is determined to be a normal generated audio signal, and the audio section and the start end and the end thereof are detected. On the other hand, in the voice section detection unit 10, the levels of the noise level voice signal c input from the output end A of the microphone 2 and the voice signal b ′ output from the output end B of the microphone 1 and input through the delay circuit 12 are compared. In comparison, when the level difference is larger than the internally determined level difference threshold, the one with the larger level, that is, the audio signal b ′ in this case, is determined as the normal uttered audio signal, and the audio Detects the section and the beginning and end. That is, the voice section detection unit 9 detects the original voiced voice of the own system uttered toward the microphone 1, but the voice section detection unit 10 detects the original voiced voice uttered toward the microphone 2 of the own system. Instead of detecting it, the uttered voice uttered toward the microphone 1 of the other system is detected as if uttered toward the microphone 2 of the own system. However, the voice section of the voice signal b ′ detected by the voice section detection unit 10 and its start and end are the voice section of the voice signal a detected by the voice section detection unit 9 and its start end due to the delay in the delay circuit 12. , Is output to the feature extraction unit h after the end. Therefore, the selection unit 7 receives the feature vector sequence g from the feature extraction unit 5 and the feature vector sequence h from the feature extraction unit 6 at the same time.
When the two signals are compared at the same time, the end of the voice signal b'added to the feature vector sequence h is necessarily delayed from the end of the voice signal a added to the feature vector sequence g. become. As a result, the comparison unit 7 selects the feature vector series g on the side whose end is detected earlier and outputs it to the recognition processing unit 8.

Similarly, when a voice is generated toward the microphone 2, the end of the voice signal c detected by the voice section detection unit 10 is the voice signal a detected by the voice section detection unit 9. Since it is earlier than the end, the selection unit 7 selects the feature vector series h from the microphone 2 side and outputs it to the recognition processing unit 8.

As described above, no matter which side of the microphone 1 and the microphone 2 the voice is uttered, the uttered voice from the microphone in which the voice is uttered can be surely selected and recognized. It is not necessary to limit one of the two microphones for voice input and the other for noise input. For example, when the microphone 1 is used for voice input, the microphone 2 is used.
Is for noise input, and when the microphone 2 is for voice input, the microphone 1 is for noise input.

[0028]

As described above, according to the present invention, 2
In a voice recognition device that performs input voice detection, it is not necessary to limit one of the two microphones for voice input and the other for noise input. For example, when one of the two microphones is for voice input, the other is for noise input, and When the other is used for voice input, one is used for noise input, and therefore both microphones can be used for voice input without distinction. As a result, the complexity of checking the microphones for voice input and noise input before use is eliminated, and even if the connection systems of both microphones are incorrect, they can be used as they are, thus eliminating the complexity of connection confirmation. It is possible to make a mistake in using the device.

[Brief description of the drawings]

FIG. 1 is a block diagram of a voice recognition device showing a first embodiment of the present invention.

FIG. 2 is a block diagram of a voice recognition device showing a second embodiment of the present invention.

FIG. 3 is a block diagram showing a first conventional example.

FIG. 4 is a block diagram showing a second example and a conventional example.

FIG. 5 is a block diagram showing a third conventional example.

[Explanation of symbols]

 1 and 2 Microphones 3 and 4 Speech section detection section 5 and 6 Feature extraction section 7 Selection section 8 Recognition processing section 9 and 10 Speech section detection section 11 and 12 Delay circuit af Speech signals g, h and i Feature vector series b ′, D ′ voice signal

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yasuo Tomooka 5-7-1, Shiba, Minato-ku, Tokyo NEC Corporation

Claims (5)

[Claims] 1. A voice recognition device for recognizing an uttered voice by performing two-input voice detection by a first and a second microphone, wherein when the first microphone is used for voice input, the second microphone is A voice recognition device, which is for noise input, and when the second microphone is for voice input, the first microphone is for noise input. 2. In a voice recognition device for recognizing an uttered voice by performing two-input voice detection by the first and second microphones, the first voice section detection means is uttered toward the first microphone. The uttered voice is input to the voice input end as a first voice signal from the first output end of the first microphone, and noise is output as a first noise signal from the second output end of the second microphone. The voice section of the voice which is input to the input end and uttered toward the first microphone is detected, the first voice signal in the voice section is cut out and output as a third voice signal, and the second voice is output. The section detecting means outputs the uttered voice when uttered toward the second microphone to the voice input end as the second voice signal from the first output end of the second and microphone. The voice section of the voice uttered toward the second microphone, which is input to the noise input end as a noise signal from the second output end of the first microphone and is applied to the second voice end. The second audio signal is cut out and output as a fourth audio signal, and the third audio signal is output.
A voice recognition device for recognizing each voiced voice uttered toward each microphone based on the voice signal and the fourth voice signal. 3. A voice recognition device for recognizing an uttered voice by performing two-input voice detection by the first and second microphones, when the first voice section detecting means is uttered toward the first microphone. The vocalized voice is input from the first output end of the first microphone to the first voice input end as a first and voice signal and the second voice of the second microphone is input.
Is input to the second voice input end through the first delay means as the first noise signal from the output end of the first voice signal, and the voice section of the voice uttered toward the first microphone is detected to detect the voice section of the voice section. The first voice signal is cut out and output as a third voice signal, and the second voice section detecting means outputs a voiced voice when the second voice section is uttered toward the second microphone. The first audio signal from the output end is the first
Of the second microphone from the second output end of the first microphone to the second voice input end through the second delay means toward the second microphone. A voice section of the uttered voice is detected, the second voice signal in the voice section is cut out and output as a fourth voice signal, and each is output based on the third voice signal and the fourth voice signal. A voice recognition device characterized by recognizing each voiced voice uttered toward each microphone. 4. A first microphone and a second microphone, which are installed close to each other, for converting vocalized voice into an electric signal, and a first electric signal output from a first output end of the first microphone. To a first input end as a first voice signal and a second electric signal output from a second output end of the second microphone to a second input end as a first noise signal. Then, only when the level of the first voice signal is higher than the level of the first noise signal, the first voice output toward the first microphone is output.
First voice detection means for detecting a voice section of the voice signal as a first voice section and cutting out the first voice signal of the first voice section and outputting it as a second voice signal, The second electric signal output from the first output end of the second microphone is input to the first input end as a third audio signal and output from the second output end of the first microphone. The second microphone is input only when the level of the third voice signal is higher than the level of the second noise signal by inputting the first electric signal to the second input terminal as the second noise signal. A voice section of the third voice signal which is uttered toward and is output as a second voice section, and the third voice signal of the second voice section is cut out to obtain a fourth voice signal. Second sound output as A first feature extraction unit that calculates a feature amount of the third audio signal from the first voice detection unit, converts the feature amount into a feature vector sequence, and outputs the feature vector sequence as a first feature vector sequence; Second feature extracting means for calculating a feature amount of the fourth voice signal from the second voice detecting means, converting the feature amount into a feature vector sequence, and outputting as a second feature vector sequence; and the first feature extracting means. From the first feature vector sequence input from the second feature extraction means, and the feature vector sequence whose end of the voice section is detected first is compared with the second feature vector sequence input from the second feature extraction means. For selecting and outputting as a third feature vector sequence, and a selector for inputting from the third feature vector sequence input from the selecting unit to the first microphone or the second microphone. A speech recognition apparatus comprising: a recognition processing unit that recognizes and processes a uttered voice that is uttered. 5. The second audio signal output from the second output end of the second microphone by the first voice detection means.
Electrical signal is input to the second input terminal as a fifth audio signal through a first delay unit having a predetermined first delay time, and is input from a first output terminal of the first microphone. The level of the first audio signal and the level of the fifth audio signal are compared to detect the audio section of the audio signal of the higher level to make the third audio section, and the third audio section. Of the audio signal is output as the third audio signal, and the second audio detecting means outputs the first electric signal output from the second output end of the first microphone to a predetermined first audio signal. A second audio signal input to the second input terminal as a sixth audio signal through a second delay means having a delay time of 2 and input from the first output terminal of the second microphone; Of the level and the sixth audio signal Comparing with the level, the voice section of the voice signal of the higher level is detected and set as the fourth voice section, and the voice signal of the third voice section is cut out and output as the fourth voice signal. The voice recognition device according to claim 4, wherein