JP7435948B2 - Sound collection device, sound collection method and sound collection program - Google Patents

Description

The present disclosure relates to a technique for collecting target sound using a plurality of microphone elements.

BACKGROUND ART Beamformers that control directivity using output signals from at least two microphone elements have been known. There is a sound collection device that uses this beamformer to suppress ambient noise and separate target sound from ambient noise. The noise suppression performance of the beamformer may deteriorate due to variations in sensitivity between at least two microphone elements.

For example, Patent Document 1 discloses a beamformer that combines a generalized sidelobe canceller (hereinafter referred to as GSC) with automatic calibration processing. In Patent Document 1, sensitivity variations among a plurality of microphones are corrected by ambient noise.

Patent No. 4734070 specification

However, with the above-mentioned conventional technology, there is a risk that the noise suppression performance in directional synthesis may deteriorate, so further improvement is required.

The present disclosure has been made to solve the above problems, and provides a technology that can improve noise suppression performance in directional synthesis and can collect target sound with a high S/N ratio. The purpose is to

A sound collection device according to an aspect of the present disclosure includes a plurality of microphone elements, a sensitivity correction unit that corrects a sensitivity difference between the plurality of microphone elements by multiplying an output signal of the plurality of microphone elements by a gain, and an utterance. a target sound detection unit that detects a person's voice as a target sound; a gain control unit that controls the gain based on a detection result of the target sound detection unit; and a gain control unit that controls the gain of the plurality of microphone elements corrected by the sensitivity correction unit. a directional synthesis unit that uses the output signal to emphasize and collect the target sound coming from a predetermined direction; is detected, the gain is updated based on the output signals of the plurality of microphone elements, and when the target sound detection section does not detect the voice of the speaker, the gain is not updated.

According to the present disclosure, noise suppression performance in directional synthesis can be improved, and target sound can be collected with a high S/N ratio.

FIG. 1 is a block diagram showing the configuration of a sound collection device in Embodiment 1 of the present disclosure. FIG. 3 is a diagram showing an example of an installation position of a microphone array in Embodiment 1 of the present disclosure. FIG. 3 is a diagram illustrating an example arrangement of microphone elements of a microphone array according to Embodiment 1 of the present disclosure. FIG. 2 is a block diagram showing the configuration of a target sound detection section of the sound collection device in Embodiment 1 of the present disclosure. FIG. 2 is a block diagram showing the configuration of a sound determination section of the sound collection device according to Embodiment 1 of the present disclosure. FIG. 2 is a block diagram showing the configuration of a sensitivity correction control section of the sound collection device in Embodiment 1 of the present disclosure. It is a flow chart for explaining operation of a sound collection device in Embodiment 1 of this indication. FIG. 2 is a block diagram showing the configuration of a sound collection device in Embodiment 2 of the present disclosure. It is a block diagram showing the composition of the target sound detection part of the sound collection device in Embodiment 2 of this indication. FIG. 3 is a block diagram showing the configuration of a target sound direction determining section of a sound collection device in Embodiment 2 of the present disclosure. FIG. 3 is a block diagram showing the configuration of a target sound direction determining section of a sound collection device in a modification of Embodiment 2 of the present disclosure.

(Findings that formed the basis of this disclosure)
As described above, in the prior art, sensitivity variations among a plurality of microphones are corrected by ambient noise.

However, when a noise source is near a microphone array consisting of multiple microphone elements, the distance difference between the noise source and each microphone element cannot be ignored, and the noise generated from the noise source is transmitted to each microphone element. This appears as a sound pressure difference at the position. If the sensitivity correction or automatic calibration of multiple microphone elements is performed due to noise generated from such a noise source near the microphone array, the sensitivity correction or automatic calibration may not be performed correctly, and the beamformer in the subsequent stage may be damaged. There was a risk that the output performance would deteriorate.

In particular, in GSC, a blocking matrix creates a noise reference signal with a blind spot of sensitivity in the direction of the target sound. However, if there are variations in sensitivity among a plurality of microphone elements, a blind spot of sensitivity cannot be formed in the direction of the target sound, and the target sound leaks into the noise reference signal. In this case, the noise reference signal into which the target sound has leaked through the subsequent adaptive noise canceling is subtracted from the output of the weighted sum beamformer, which may distort the target sound of the output signal. In order to prevent the target sound from leaking into the noise reference signal, it is necessary to equalize the sensitivity of at least a plurality of microphone elements.

In order to solve the above problems, a sound collection device according to one aspect of the present disclosure includes a plurality of microphone elements and a sensitivity difference between the plurality of microphone elements by multiplying an output signal of the plurality of microphone elements by a gain. a target sound detection unit that detects the speaker's voice as the target sound, a gain control unit that controls the gain based on the detection result of the target sound detection unit, and the sensitivity correction unit. a directional synthesis section that emphasizes and collects the target sound coming from a predetermined direction using the corrected output signals of the plurality of microphone elements; If the detection unit detects the voice of the speaker, the gain is updated based on the output signals of the plurality of microphone elements, and if the target sound detection unit does not detect the voice of the speaker, The gain is not updated.

According to this configuration, the sensitivity difference between the plurality of microphone elements is corrected by multiplying the output signals of the plurality of microphone elements by the gain. At this time, if the speaker's voice is detected, the gain is updated based on the output signals of the plurality of microphone elements, and if the speaker's voice is not detected, the gain is not updated. Then, using the output signals of the plurality of microphone elements whose sensitivity differences have been corrected, the target sound coming from a predetermined direction is emphasized and collected.

Therefore, when the speaker's voice, which is the target sound, is detected, the gain for correcting the sensitivity difference between multiple microphone elements is updated, so the sensitivity difference between the multiple microphone elements for the target sound is corrected. In the subsequent directional synthesis, it is possible to reduce the amount of the target sound leaking into the noise reference signal that has a blind spot in sensitivity in the direction of the target sound. As a result, the noise suppression performance in directional synthesis can be improved, and the target sound can be collected with a high S/N ratio.

Further, in the above sound collection device, the target sound detection unit determines whether the output signal of one of the plurality of microphone elements is the voice or a non-voice other than the voice. It may also include a voice determination section.

According to this configuration, the voice of the speaker can be easily detected by determining whether the output signal of one of the plurality of microphone elements is voice or non-voice. .

Further, in the above sound collection device, the target sound detection section includes a first extraction section that extracts a signal in a specific band from the output signal of the one microphone element, and the sound determination section includes the first extraction section. It may be determined whether the signal extracted by the unit is the voice or the non-voice.

According to this configuration, it is determined whether the signal in a specific band extracted from the output signal of one microphone element is voice or non-voice, so the voice of the speaker can be detected with higher accuracy. can be detected.

Further, in the above-mentioned sound collection device, the purpose of the target sound detection unit is to determine whether or not the target sound is coming from a predetermined target sound direction using the output signals of the plurality of microphone elements. A sound direction determination unit and the target sound direction determination unit determine that the target sound is coming from the target sound direction, and the audio determination unit determines that the output signal of the one microphone element is the voice. The target sound determination unit may also include a target sound determining unit that determines that the target sound has been detected when the target sound is determined.

If only the determination is made as to whether or not speech has been detected, there is a risk that it will be determined that speech has been detected even if speech is made from a direction other than the direction of the target sound. On the other hand, according to the above configuration, only when a voice is detected and the target sound is coming from the direction of the target sound, it is determined that the target sound has been detected and the gain is updated. Sensitivity differences can be corrected using sound.

Further, in the above sound collection device, the target sound detection unit includes a second extraction unit that extracts a signal in a specific band from the output signals of the plurality of microphone elements, and the target sound direction determination unit includes a second extraction unit that extracts a signal in a specific band from the output signals of the plurality of microphone elements, and It may be determined whether or not the target sound is coming from the target sound direction with respect to the signal extracted by the second extraction unit.

According to this configuration, it is determined whether or not the target sound is coming from the target sound direction with respect to the signal of a specific band extracted from the output signals of the plurality of microphone elements, so that the target sound can be heard with higher accuracy. It can be determined whether the target sound is coming from the direction.

Further, in the above sound collection device, the target sound direction determination unit includes a direction estimation unit that estimates a direction in which the target sound arrives using a phase difference between output signals of the plurality of microphone elements; and a direction determination unit that determines whether the direction estimated by the unit is the predetermined target sound direction.

The direction in which the target sound arrives can be easily estimated by using the phase difference between the output signals of the plurality of microphone elements. Therefore, whether or not the target sound is coming from the target sound direction can be easily determined based on the estimation result of the direction in which the target sound is coming, if the target sound direction is known in advance.

Further, in the above-mentioned sound collection device, the target sound direction determination unit is configured to form a directivity in the target sound direction by emphasizing a signal in the target sound direction using the output signals of the plurality of microphone elements. a second directivity synthesis section that uses the output signals of the plurality of microphone elements to form a blind spot of sensitivity in the direction of the target sound, and an output signal from the first directionality synthesis section. and a level comparison determination unit that compares the level with the output level of the output signal from the second directional synthesis unit and determines whether or not the target sound is coming from the target sound direction. .

When the target sound is coming from the target sound direction, the output signal level from the first directional synthesis section is higher than the output signal level from the second directional synthesis section. Therefore, if the output signal level from the first directional synthesis section is higher than the output signal level from the second directional synthesis section, it can be determined that the target sound is coming from the direction of the target sound. On the other hand, when the target sound is not arriving from the target sound direction, the output signals of the first directional synthesis section and the second directional synthesis section include only peripheral noise. Therefore, the output signal level from the first directivity combining section is approximately equal to the output signal level from the second directivity combining section, or is smaller than the output signal level from the second directivity combining section. Therefore, when the output signal level from the first directional synthesis section is equal to or lower than the output signal level from the second directional synthesis section, it can be determined that the target sound is not coming from the direction of the target sound.

Further, in the above sound collection device, the gain control section includes a level detection section that detects the output level of the output signal of each of the plurality of microphone elements, and a level detection section that detects the voice of the speaker by the target sound detection section. a time average level calculation unit that calculates a time average level of each output level detected by the level detection unit; and a time average level calculation unit that updates the gain from the time average level calculated by the time average level calculation unit. A correction gain calculation unit that calculates a correction gain may also be included.

According to this configuration, when the voice of the speaker is detected, the time average level of the output level of the output signal of each of the plurality of microphone elements is calculated. Then, a correction gain that updates the gain is calculated from the calculated time average level, so the sensitivity difference between the plurality of microphone elements is corrected by multiplying the output signals of the plurality of microphone elements by the calculated correction gain. be able to.

Further, in the above-mentioned sound collection device, the correction gain calculation unit may calculate the time average level of one microphone element determined in advance among the plurality of microphone elements as a reference, and calculate The correction gain of the other microphone element may be calculated so that the time average level of the microphone element is the same as the time average level of the one microphone element.

According to this configuration, the sensitivity difference between the plurality of microphone elements is adjusted so that the output level of the other microphone elements is equal to the predetermined output level of one of the plurality of microphone elements. Can be corrected.

Further, in the above-mentioned sound collection device, the correction gain calculation unit calculates the time average level of at least two predetermined microphone elements of the plurality of microphone elements as a reference. The correction gains of the plurality of microphone elements may be calculated such that the time average level of the element is the same as the average value of the time average levels of the at least two microphone elements.

According to this configuration, between the plurality of microphone elements, the output levels of the plurality of microphone elements are equalized with respect to the average value of the output levels of at least two microphone elements determined in advance among the plurality of microphone elements. The difference in sensitivity can be corrected.

Further, in the above sound collection device, the gain control section includes a third extraction section that extracts a signal in a specific band from the output signal of each of the plurality of microphone elements, and the level detection section includes the third extraction section. The output level of each signal extracted by the section may be detected.

According to this configuration, since the output level of each signal in a specific band extracted from the output signals of each of the plurality of microphone elements is detected, the influence of noise other than the target sound can be reduced.

Moreover, in the above-mentioned sound collection device, the specific band may be a band from 200Hz to 500Hz.

According to this configuration, the output level of each signal in the band from 200 Hz to 500 Hz extracted from the output signal of each of the plurality of microphone elements is detected. Therefore, by removing low-frequency noise of 200 Hz or less, the influence of low-frequency noise can be reduced. Furthermore, by removing the band of 500 Hz or more, the sound is limited to sounds with a wavelength sufficiently longer than the size of the microphone array, and the difference in sound pressure depending on the position of the microphone elements forming the microphone array is reduced. This enables accurate sensitivity correction.

In a sound collection method according to another aspect of the present disclosure, the computer corrects the sensitivity difference between the plurality of microphone elements by multiplying the output signals of the plurality of microphone elements by a gain, and uses the voice of the speaker as the target sound. detecting the target sound, controlling the gain based on the detection result of the target sound, and using the corrected output signals of the plurality of microphone elements to emphasize and collect the target sound coming from a predetermined direction. , in controlling the gain, if the voice of the speaker is detected, the gain is updated based on the output signals of the plurality of microphone elements, and if the voice of the speaker is not detected, the gain is updated. is not updated.

According to this configuration, the sensitivity difference between the plurality of microphone elements is corrected by multiplying the output signals of the plurality of microphone elements by the gain. At this time, if the speaker's voice is detected, the gain is updated based on the output signals of the plurality of microphone elements, and if the speaker's voice is not detected, the gain is not updated. Then, using the output signals of the plurality of microphone elements whose sensitivity differences have been corrected, the target sound coming from a predetermined direction is emphasized and collected.

Therefore, when the speaker's voice, which is the target sound, is detected, the gain for correcting the sensitivity difference between multiple microphone elements is updated, so the sensitivity difference between the multiple microphone elements for the target sound is corrected. In the subsequent directional synthesis, it is possible to reduce the amount of the target sound leaking into the noise reference signal that has a blind spot in sensitivity in the direction of the target sound. As a result, the noise suppression performance in directional synthesis can be improved, and the target sound can be collected with a high S/N ratio.

A sound collection program according to another aspect of the present disclosure includes a sensitivity correction unit that corrects sensitivity differences between the plurality of microphone elements by multiplying output signals of the plurality of microphone elements by a gain, and converts the voice of a speaker into a target sound. a target sound detection unit that detects the target sound as a target sound, a gain control unit that controls the gain based on the detection result of the target sound detection unit, and the output signals of the plurality of microphone elements corrected by the sensitivity correction unit. , the computer functions as a directional synthesis unit that emphasizes and collects the target sound coming from a predetermined direction, and the gain control unit controls when the voice of the speaker is detected by the target sound detection unit. , the gain is updated based on the output signals of the plurality of microphone elements, and the gain is not updated when the target sound detection unit does not detect the voice of the speaker.

According to this configuration, the sensitivity difference between the plurality of microphone elements is corrected by multiplying the output signals of the plurality of microphone elements by the gain. At this time, if the speaker's voice is detected, the gain is updated based on the output signals of the plurality of microphone elements, and if the speaker's voice is not detected, the gain is not updated. Then, using the output signals of the plurality of microphone elements whose sensitivity differences have been corrected, the target sound coming from a predetermined direction is emphasized and collected.

Therefore, when the speaker's voice, which is the target sound, is detected, the gain for correcting the sensitivity difference between multiple microphone elements is updated, so the sensitivity difference between the multiple microphone elements for the target sound is corrected. In the subsequent directional synthesis, it is possible to reduce the amount of the target sound leaking into the noise reference signal that has a blind spot in sensitivity in the direction of the target sound. As a result, the noise suppression performance in directional synthesis can be improved, and the target sound can be collected with a high S/N ratio.

Embodiments of the present disclosure will be described in detail below with reference to the drawings. Note that the following embodiments are examples that embody the present disclosure, and do not limit the technical scope of the present disclosure.

(Embodiment 1)
FIG. 1 is a block diagram showing the configuration of a sound collection device according to Embodiment 1 of the present disclosure.

The sound collection device 101 shown in FIG. 1 includes a microphone array 1, a sensitivity correction section 2, a target sound detection section 3, a sensitivity correction control section (gain control section) 4, and a directivity synthesis section 5.

The microphone array 1 includes n microphone elements 11, 12, . . . , 1n (n is a natural number) that convert acoustic signals into electrical signals. Microphone array 1 includes a plurality of microphone elements.

FIG. 2 is a diagram illustrating an example of the installation position of the microphone array according to Embodiment 1 of the present disclosure, and FIG. 3 is a diagram illustrating an example of the arrangement of microphone elements of the microphone array according to Embodiment 1 of the present disclosure. .

As shown in FIG. 2, the microphone array 1 in the first embodiment is arranged near a display 201 in the vehicle. Display 201 is a component of a car navigation system. Further, below the display 201, an air outlet 202 of an air conditioner is provided. Cooled air or warmed air is output from the air outlet 202 .

Further, the microphone array 1 shown in FIG. 3 includes, for example, four microphone elements 11, 12, 13, and 14. Microphone elements 11, 12, 13, and 14 are arranged at each of the four corners of a rectangular substrate. The horizontal distance between the microphone elements 11 and 12 arranged at the bottom of the substrate is, for example, 2 cm. Further, the horizontal interval between the microphone elements 13 and 14 arranged on the upper part of the substrate is, for example, 2 cm. Further, the vertical distance between the microphone elements 11 and 13 is, for example, 2 cm, and the vertical distance between the microphone elements 12 and 14 is, for example, 2 cm.

The distance between the microphone array 1 and the air outlet 202 is, for example, 2 cm. The microphone array 1 acquires the voice of a speaker sitting in a driver's seat as a target sound. At this time, the sound of the air output from the air outlet 202 is included as noise in the target sound. The distance between the microphone element 13 closest to the air outlet 202 and the air outlet 202 is 2 cm, and the distance between the microphone element 11 furthest from the air outlet 202 and the air outlet 202 is 4 cm. The distance between the microphone element 11 and the air outlet 202 is twice the distance between the microphone element 13 and the air outlet 202.

In this case, the distance difference between the air outlet 202, which is a noise source, and each microphone element 11, 13 cannot be ignored, and the noise generated from the air outlet 202 is equal to the sound pressure difference at the position of each microphone element 11, 13. It appears. If the sensitivity of the microphone elements 11 to 14 is corrected using noise generated from a noise source near the microphone array 1, the sensitivity may not be corrected correctly, and the directivity synthesis section (beam There was a risk that the output performance of the former (former) would deteriorate. Therefore, the sound collection device 101 in the first embodiment corrects the sensitivity of the microphone elements 11 to 14 using the target sound.

Note that the number of microphone elements included in the microphone array 1 is not limited to four. Furthermore, the arrangement positions of the plurality of microphone elements are not limited to the arrangement positions shown in FIG. 3.

An output signal from one of the microphone elements 11, 12, . . . , 1n is input to the target sound detection section 3. Further, each output signal of the microphone elements 11, 12, . . . , 1n is input to the sensitivity correction section 2 and the sensitivity correction control section 4.

The sensitivity correction section 2 corrects the sensitivity difference between the plurality of microphone elements 11, 12, . . . , 1n by multiplying the output signals of the plurality of microphone elements 11, 12, . . . , 1n by gain. The sensitivity correction unit 2 corrects variations in sensitivity of each microphone element 11, 12, . . . , 1n by multiplying the output signal of each microphone element 11, 12, . . . , 1n by a specified gain. The sensitivity correction section 2 equalizes the sensitivity among the plurality of microphone elements 11, 12, . . . , 1n.

The target sound detection unit 3 detects the speaker's voice as the target sound. The target sound detection unit 3 acquires the output signal of one of the microphone elements 11, 12, . Note that in the first embodiment, the audio determination unit 32 uses the output signal of the microphone element 11 to detect the presence or absence of a target sound, but the present disclosure is not particularly limited thereto. The audio determination unit 32 may detect the presence or absence of the target sound using the output signal of any one of the microphone elements 11, 12, . . . , 1n.

Note that the configuration of the target sound detection section 3 will be explained in more detail using FIGS. 4 and 5.

The sensitivity correction control section 4 controls the gain based on the detection result of the target sound detection section 3. The sensitivity correction control section 4 acquires the output signal of each microphone element 11, 12, . , 12, . . . , 1n.

When the target sound detection unit 3 detects the speaker's voice, the sensitivity correction control unit 4 updates the gain based on the output signals of the plurality of microphone elements, and the target sound detection unit 3 detects the speaker's voice. If not, do not update the gain. Note that the configuration of the sensitivity correction control section 4 will be explained in more detail using FIG. 6.

A directional synthesis section (beamformer) 5 uses the output signals of the plurality of microphone elements corrected by the sensitivity correction section 2 to emphasize and collect the target sound coming from a predetermined direction. The directivity synthesis section 5 acquires the output signals of the microphone elements 11, 12, . . . , 1n corrected by the sensitivity correction section 2, and improves the S/N ratio of the target sound.

Next, the configuration of the target sound detection section 3 shown in FIG. 1 will be further explained.

FIG. 4 is a block diagram showing the configuration of the target sound detection section of the sound collection device according to Embodiment 1 of the present disclosure.

The target sound detection section 3 shown in FIG. 4 includes a bandpass filter section (first extraction section) 31 and a voice determination section 32.

The bandpass filter section 31 extracts a signal in a specific band from the output signal of one microphone element 11 among the plurality of microphone elements 11, 12, . . . , 1n. The bandpass filter section 31 extracts a signal in a band from 200 Hz to 500 Hz, for example, from the output signal of the microphone element 11. The bandpass filter section 31 extracts a signal in a band from the output signal of the microphone element 11 in which the voice uttered by a person can be extracted.

The audio determination unit 32 determines whether the output signal of one microphone element 11 among the plurality of microphone elements 11, 12, . . . , 1n is audio or non-audio other than audio. The voice determination section 32 determines whether the signal extracted by the bandpass filter section 31 is voice or non-voice.

Next, the configuration of the voice determining section 32 shown in FIG. 4 will be further explained.

FIG. 5 is a block diagram showing the configuration of the audio determination section of the sound collection device according to Embodiment 1 of the present disclosure.

The voice determination section 32 includes a level detection section 321, a noise level detection section 322, a comparison section 323, a time-frequency conversion section 324, a voice feature amount extraction section 325, and a determination section 326.

The level detection section 321 detects the signal level of the output signal of the microphone element 11.

The noise level detection section 322 detects the noise level by holding the minimum value of the signal level detected by the level detection section 321.

The comparison unit 323 compares the output of the level detection unit 321 and the output of the noise level detection unit 322 to determine the presence or absence of audio at the waveform level. For example, the comparison unit 323 sets a value twice the noise level detected by the noise level detection unit 322 as the threshold value. Then, the comparison unit 323 determines whether the signal level detected by the level detection unit 321 is equal to or higher than the threshold value. The comparison unit 323 determines that the output signal of the microphone element 11 includes audio when the signal level detected by the level detection unit 321 is equal to or higher than the threshold value. On the other hand, when the signal level detected by the level detection section 321 is smaller than the threshold value, the comparison section 323 determines that the output signal of the microphone element 11 does not include audio.

The time-frequency converter 324 converts the time domain output signal of the microphone element 11 into a frequency domain output signal.

The audio feature extraction unit 325 extracts audio features from the frequency domain output signal. The audio feature amount is a feature amount indicating audio. The audio feature extracting unit 325 uses a method of extracting audio features using audio pitch as shown in Japanese Patent No. 5450298, or a method of extracting audio features using audio pitch as shown in Japanese Patent No. 3849116. Alternatively, a method may be used in which the voice feature is extracted using the voice feature as a quantity. When the sound collection device 101 is mounted on a vehicle, the microphone array 1 is installed around a display 201 embedded in a console, as shown in FIG. Therefore, the noise source is the air outlet 202 of the air conditioner. In this case, since the noise spectrum is relatively monotonous, the audio feature extraction unit 325 may extract the alternating current component of the amplitude spectrum or the ratio between the peak and dip of the amplitude spectrum as the audio feature. This makes it possible to distinguish between noise and voice generated from the air outlet 202 of the air conditioner.

If the comparison unit 323 determines that the output signal of the microphone element 11 includes voice, and the voice feature extraction unit 325 extracts the voice feature from the output signal of the microphone element 11, the determination unit 326 determines whether the microphone element It is determined that the output signal of No. 11 is audio. On the other hand, if the comparing unit 323 determines that the output signal of the microphone element 11 does not include audio, or the audio feature extracting unit 325 does not extract the audio feature from the output signal of the microphone element 11, In this case, it is determined that the output signal of the microphone element 11 is non-speech. The determination unit 326 outputs a determination result signal Odet(j) indicating either voice or non-voice to the sensitivity correction control unit 4. Note that j indicates a sample number corresponding to time.

As a result, when the target sound detection unit 3 determines that the output signal of the microphone element 11 is voice, it outputs the determination result signal Odet(j)=1, and when the output signal of the microphone element 11 is determined to be non-voice. If it is determined, a determination result signal Odet(j)=0 is output.

Next, the configuration of the sensitivity correction control section 4 shown in FIG. 1 will be further explained.

FIG. 6 is a block diagram showing the configuration of the sensitivity correction control section of the sound pickup device according to Embodiment 1 of the present disclosure.

The sensitivity correction control section 4 includes first to n band pass filter sections (third extraction section) 411 to 41n, first to n level detection sections 421 to 42n, and first to nth average level calculation sections (time average level calculation section). ) 431 to 43n and a correction gain calculation unit 44. The first to nth bandpass filter sections 411 to 41n, the first to nth level detection sections 421 to 42n, and the first to nth average level calculation sections 431 to 43n are provided according to the number of microphone elements 11 to 1n, respectively. For example, the output signal x(1,j) of the microphone element 11 is input to the first band pass filter section 411.

The first to nth bandpass filter sections 411 to 41n extract signals in specific bands from the output signals of the plurality of microphone elements 11 to 1n, respectively. Note that the specific band is a band from 200Hz to 500Hz.

The first to nth level detection units 421 to 42n detect the output levels of the output signals of the plurality of microphone elements 11 to 1n, respectively.

The first to nth level detection units 421 to 42n calculate the output level Lx (i, j) of the output signal x (i, j) of each microphone element using the following general amplitude smoothing formula (1). To detect.

Lx(i,j)=beta1・|x(i,j)|+(1−beta1)・Lx(i,j−1)...(1)

In equation (1), i indicates a microphone element number, and j indicates a sample number corresponding to time. Furthermore, in equation (1), beta1 indicates a weighting coefficient and is a parameter that determines the speed of averaging.

Furthermore, in the first embodiment, the output signal xbpf(i,j) that has passed through the first to n band pass filter sections 411 to 41n is input to the first to n level detection sections 421 to 42n. Therefore, the first to nth level detection units 421 to 42n detect the output level Lx(i,j) of the output signal xbpf(i,j) of each microphone element extracted by the first to nth bandpass filter units 411 to 41n. is detected using the following general amplitude smoothing equation (2).

Lx(i,j)=beta1・|xbp(i,j)|+(1−beta1)・Lx(i,j−1)...(2)

The first to nth average level calculation units 431 to 43n calculate each output level Lx(i , j) is calculated.

The first to nth average level calculation units 431 to 43n calculate the output level Lx(i, j ) (time average level Avex(i,j)) is calculated using the following equation (3). In addition, the first to nth average level calculation units 431 to 43n calculate the time average level Avex (i, j) as follows during a period in which the target sound is not detected by the target sound detection unit 3 (judgment result signal Odet (j) = 0). Calculated using equation (4). That is, when the target sound detection unit 3 does not detect the speaker's voice, the first to nth average level calculation units 431 to 43n use the previously calculated time average level Avex (i, j−1) this time. is calculated as the time average level Avex(i,j).

Avex(i,j)=beta2・|Lx(i,j)|+(1−beta2)・Avex(i,j−1) if Odet(j)=1...(3)

Avex(i,j)=Avex(i,j-1) if Odet(j)=0...(4)

In equations (3) and (4), i indicates a microphone element number, and j indicates a sample number corresponding to time. Furthermore, in equation (3), beta2 is a weighting coefficient and is a parameter that determines the speed of averaging. Also, beta1>>beta2. For example, if the sampling frequency is 16 kHz, beta1 is set to 0.000625 to be the average level over 100 msec, and beta2 is set to 0.0000125 to be the average level over 5 seconds. By using a long-term average level as the average signal level used for sensitivity correction of the microphone element, it is possible to accurately calculate the sensitivity correction gain.

The correction gain calculation unit 44 calculates a sensitivity correction gain with the gain updated from the time average level calculated by the first to nth average level calculation units 431 to 43n.

The correction gain calculation unit 44 calculates the level of the microphone elements 12 to 1n other than the one microphone element 11 based on the time average level of one microphone element 11 determined in advance among the plurality of microphone elements 11 to 1n. The sensitivity correction gains of the other microphone elements 12 to 1n are calculated so that the time average level is the same as the time average level of one microphone element 11. That is, the correction gain calculation unit 44 calculates the time average level Avex(i,j) of each of the microphone elements 11 to 1n calculated by the first to nth average level calculation units 431 to 43n, and the time average level Avex of the microphone element 11. (1, j), the sensitivity correction gain G(i, j) is calculated by the following equation (5).

G(i,j)=Avex(1,j)/Avex(i,j)...(5)

When the sensitivity correction gain of the above equation (5) is used, sensitivity correction is performed so that the output levels of the other microphone elements 12 to 1n are equalized with respect to the microphone element 11.

Note that in the above equation (5), the correction gain calculation unit 44 calculates the sensitivity correction gain based on the predetermined time average level of one microphone element 11, but the present disclosure particularly focuses on this. Not limited. The correction gain calculation unit 44 may calculate the sensitivity correction gain based on the time average level of another microphone element different from the microphone element 11.

Further, the correction gain calculating unit 44 calculates the time average level of the plurality of microphone elements 11 to 1n based on the average value of the time average level of at least two microphone elements determined in advance among the plurality of microphone elements 11 to 1n. The sensitivity correction gains of the plurality of microphone elements 11 to 1n may be calculated so that the level is the same as the average value of the time-averaged levels of at least two microphone elements. That is, the correction gain calculation unit 44 calculates the time average level Avex(i, j) of each microphone element 11 to 1n calculated by the first to nth average level calculation units 431 to 43n, and the time average level Avex(i, j ) may be used to calculate the sensitivity correction gain G(i, j) using the following equation (6).

G(i,j)={Avex(1,j)+Avex(2,j)+...+Avex(n,j)}/n/Avex(i,j)...(6)

Note that in the above equation (6), the correction gain calculation unit 44 calculates the sensitivity correction gain based on the average value of the time average levels of all the microphone elements 11 to 1n among the microphone elements 11 to 1n. However, the present disclosure is not particularly limited thereto. The correction gain calculation unit 44 may calculate the sensitivity correction gain based on the average value of the time average levels of at least two of the microphone elements 11 to 1n.

The sensitivity correction section 2 applies the sensitivity correction gain G(i, j) corresponding to each microphone element 11 to 1n calculated by the sensitivity correction control section 4 to the output signal x(i, j) of each microphone element 11 to 1n. Sensitivity correction is performed by multiplying.

The directivity synthesis unit 5 uses the output signals G(i,j) and x(i,j) corrected by the sensitivity correction unit 2 to perform directionality synthesis (beamforming) by GSC shown in Patent Document 1. . Further, the directivity combining unit 5 may perform beamforming using a beamforming process other than GSC, for example, an existing beamforming process such as the Maximum Likelihood method or the Minimum Variance method.

Next, the operation of the sound collection device 101 in Embodiment 1 of the present disclosure will be described.

FIG. 7 is a flowchart for explaining the operation of the sound collection device in Embodiment 1 of the present disclosure.

First, in step S1, the target sound detection unit 3 acquires an output signal from the microphone element 11, and acquires output signals from the sensitivity correction unit 2, the sensitivity correction control unit 4, and each of the microphone elements 11 to 1n.

Next, in step S2, the target sound detection unit 3 determines whether the target sound (voice) is detected from the output signal of the microphone element 11. The target sound detection unit 3 outputs a determination result signal indicating whether or not the target sound has been detected from the output signal of the microphone element 11 to the sensitivity correction control unit 4.

Here, if it is determined that the target sound is detected from the output signal of the microphone element 11 (YES in step S2), in step S3, the sensitivity correction control unit 4 to update the sensitivity correction gain.

On the other hand, if it is determined that the target sound is not detected from the output signal of the microphone element 11 (NO in step S2), the process proceeds to step S4 without updating the sensitivity correction gain.

Next, in step S4, the sensitivity correction section 2 corrects the sensitivity difference between each microphone element by multiplying the output signal of each microphone element 11 to 1n by a sensitivity correction gain.

Next, in step S5, the directivity synthesis unit 5 uses the output signals of the microphone elements 11 to 1n corrected by the sensitivity correction unit 2 to synthesize the directivity. By combining the directivity, the target sound coming from a predetermined direction is emphasized and collected.

As described above, the sensitivity difference between the plurality of microphone elements 11 to 1n is corrected by multiplying the output signals of the plurality of microphone elements 11 to 1n by a gain. At this time, if the speaker's voice is detected, the gain is updated based on the output signals of the plurality of microphone elements 11 to 1n, and if the speaker's voice is not detected, the gain is not updated. Then, using the output signals of the plurality of microphone elements 11 to 1n whose sensitivity differences have been corrected, the target sound coming from a predetermined direction is emphasized and collected.

Therefore, when the target sound of the speaker's voice is detected, the gain for correcting the sensitivity difference between the plurality of microphone elements 11 to 1n is updated, so that the plurality of microphone elements 11 to 1n for the target sound It is possible to correct the difference in sensitivity between the two, and in the subsequent directional synthesis, it is possible to reduce the amount of the target sound leaking into the noise reference signal that has a blind spot of sensitivity in the direction of the target sound. As a result, the noise suppression performance in directional synthesis can be improved, and the target sound can be collected with a high S/N ratio.

(Embodiment 2)
In the first embodiment described above, the target sound detection unit 3 determines whether the output signal of one microphone element is voice or non-voice. In contrast, in the second embodiment, the target sound detection unit further determines whether or not the target sound is coming from a predetermined target sound direction using the output signals of the plurality of microphone elements.

FIG. 8 is a block diagram showing the configuration of a sound collection device in Embodiment 2 of the present disclosure.

The sound collection device 102 shown in FIG. 8 includes a microphone array 1, a sensitivity correction section 2, a sensitivity correction control section 4, a directional synthesis section 5, and a target sound detection section 6. The difference from the sound collection device 101 of the first embodiment is that output signals from a plurality of microphone elements 11, 12, . . . , 1n are input to the target sound detection section 6. Note that in the second embodiment, the same components as in the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted.

FIG. 9 is a block diagram showing the configuration of the target sound detection section of the sound collection device in Embodiment 2 of the present disclosure.

The target sound detection unit 6 shown in FIG. 9 includes a bandpass filter unit 31, a voice determination unit 32, a bandpass filter unit (second extraction unit) 63, a target sound direction determination unit 64, and a target sound determination unit 65. In contrast to the target sound detection unit 3 of the first embodiment, the target sound detection unit 6 of the second embodiment includes a bandpass filter unit 63, a target sound direction determination unit 64, and a target sound determination unit 65. .

The bandpass filter section 63 extracts signals in a specific band from the output signals of the plurality of microphone elements. The bandpass filter section 63 extracts signals in a band of, for example, 200 Hz to 500 Hz from the output signals of the microphone elements 11 to 1n.

The target sound direction determination unit 64 determines whether or not the target sound is coming from a predetermined target sound direction using the output signals of the plurality of microphone elements. The target sound direction determination unit 64 determines whether or not the target sound is coming from the target sound direction with respect to the signal extracted by the bandpass filter unit 63. Here, when the sound collection device 102 placed in the vehicle collects the voice uttered by the driver, the angle at which the voice uttered by the driver enters the microphone array 1 is determined in advance. Therefore, the target sound direction determination unit 64 stores the incident angle of the uttered sound in advance. Note that the configuration of the target sound direction determining section 64 will be explained in more detail using FIGS. 10 and 11.

The target sound determination section 65 determines the presence or absence of the target sound using the two determination results from the voice determination section 32 and the target sound direction determination section 64. The target sound determination unit 65 determines that the target sound is coming from the target sound direction by the target sound direction determination unit 64, and the audio determination unit 32 determines that the output signal of one microphone element is voice. If so, it is determined that the target sound has been detected. Further, when the target sound direction determining unit 64 determines that the target sound is not coming from the target sound direction, or the audio determining unit 32 determines that the output signal of one microphone element is not a voice. If it is determined that the target sound is not detected, it is determined that the target sound is not detected.

Next, the configuration of the target sound direction determining section 64 shown in FIG. 9 will be further explained.

FIG. 10 is a block diagram showing the configuration of a target sound direction determining section of a sound collection device in Embodiment 2 of the present disclosure. Note that in FIG. 10, for convenience of explanation, an example will be described in which output signals from two microphone elements 11 and 12 are input to the target sound direction determining section 64.

The target sound direction determination section 64 includes a delay-sum directivity synthesis section (delay-sum beamformer) (first directionality synthesis section) 641, and a tilt-type directivity synthesis section (gradient-type beamformer) (second directionality synthesis section). 642, a target sound level detection section 643, a non-target sound level detection section 644, and a level comparison determination section 645.

The delay-sum directivity synthesis unit 641 forms directivity in the direction of the target sound by emphasizing the signal in the direction of the target sound using the output signals of the plurality of microphone elements 11 to 1n. The delay-sum directivity synthesis section 641 has high directivity sensitivity in the direction of the target sound. A directional characteristic 6411 shown in FIG. 10 indicates a directional characteristic of the delay-sum directivity synthesis section 641. The directional characteristic 6411 of the delay-sum directional synthesis unit 641 has directivity in the direction of the target sound, and emphasizes the signal in the direction of the target sound.

The delay-sum directivity synthesis unit 641 converts the output signal from the microphone element 11 into a path difference Δ(Δ =dsinθ). Then, the delay-sum directivity synthesis section 641 adds the delayed output signal from the microphone element 11 and the output signal from the microphone element 12. Note that the distance d and the incident angle θ are stored in advance in a memory (not shown).

The gradient-type directivity synthesis section 642 forms a sensitivity blind spot in the direction of the target sound using the output signals of the plurality of microphone elements 11 and 12. A directional characteristic 6421 shown in FIG. 10 indicates a directional characteristic of the gradient-type directivity synthesis section 642. The directional characteristic 6421 of the gradient-type directional synthesis unit 642 has a blind spot in the direction of the target sound, and emphasizes the signal (noise) in the direction perpendicular to the direction of the target sound.

The gradient-type directivity synthesis unit 642 converts the output signal from the microphone element 11 into a path difference Δ, where d is the distance between the microphone element 11 and the microphone element 12, and θ is the incident angle of sound from the target sound direction. (Δ=dsinθ). Then, the gradient-type directivity synthesis unit 642 subtracts the output signal from the microphone element 12 from the delayed output signal from the microphone element 11. Note that the distance d and the incident angle θ are stored in advance.

The target sound level detection section 643 detects the output signal level of the delay-sum directivity synthesis section 641.

The non-target sound level detection section 644 detects the output signal level of the gradient-type directivity synthesis section 642.

The level comparison/judgment unit 645 compares the output level of the output signal from the delay sum directivity synthesis unit 641 and the output level of the output signal from the gradient type directivity synthesis unit 642, and determines whether the target sound is coming from the direction of the target sound. Determine whether or not. The level comparison and determination section 645 compares the output signal level detected by the target sound level detection section 643 and the output signal level detected by the non-target sound level detection section 644, and determines whether the target sound is coming from the direction of the target sound. Determine whether or not the

The delay-sum directivity synthesis section 641 has directivity in the direction of the target sound. Therefore, the speaker's voice, which is the target sound, is included in the output of the delay-sum directional synthesis section 641. On the other hand, the gradient-type directivity synthesis section 642 has a blind spot in the direction of the target sound. Therefore, the speaker's voice, which is the target sound, is hardly included in the output of the gradient-type directional synthesis unit 642. Therefore, when the target sound is coming from the target sound direction, the output signal level detected by the target sound level detection section 643 becomes high, and the output signal level detected by the non-target sound level detection section 644 becomes low. When the output signal level (target sound level) detected by the target sound level detection unit 643 is higher than the output signal level (non-target sound level) detected by the non-target sound level detection unit 644, the level comparison determination unit 645 determines that It is determined that the target sound is coming from the direction of the target sound.

On the other hand, when the target sound does not arrive from the target sound direction, the outputs of the delay-sum directivity synthesis section 641 and the gradient-type directivity synthesis section 642 include only peripheral noise. Therefore, the output signal level detected by the target sound level detection section 643 is approximately equal to the output signal level detected by the non-target sound level detection section 644, or the output signal level detected by the non-target sound level detection section 644 is approximately equal to the output signal level detected by the non-target sound level detection section 644. smaller than the level. When the output signal level (target sound level) detected by the target sound level detection unit 643 is equal to or lower than the output signal level (non-target sound level) detected by the non-target sound level detection unit 644, the level comparison determination unit 645 determines that , it is determined that the target sound is not coming from the target sound direction.

In the first embodiment, when a voice is detected, it is determined that the target sound has been detected. Therefore, even if speech is made from a direction other than the direction of the target sound, it is determined that the target sound has been detected, and sensitivity correction is performed. It will be done. On the other hand, in the second embodiment, it is determined that the target sound has been detected only when a voice is detected and the target sound is coming from the direction of the target sound. Therefore, the sound collection device 102 of the second embodiment can perform sensitivity correction using the target sound with higher accuracy than the sound collection device 101 of the first embodiment.

Next, the configuration of the target sound direction determining section in a modification of the second embodiment will be further described.

FIG. 11 is a block diagram showing the configuration of a target sound direction determining section of a sound collection device in a modification of Embodiment 2 of the present disclosure. Note that in FIG. 11, for convenience of explanation, an example will be described in which output signals from two microphone elements 11 and 12 are input to the target sound direction determining section 64A. Further, the target sound detection unit 6 shown in FIG. 9 includes a target sound direction determination unit 64A shown in FIG. 11 instead of the target sound direction determination unit 64 shown in FIG.

The target sound direction determination unit 64A includes a target sound direction estimation unit (direction estimation unit) 646 and a direction determination unit 647.

The target sound direction estimation unit 646 estimates the direction in which the target sound arrives, using the phase difference between the output signals of the plurality of microphone elements. A memory (not shown) stores the distance d between the microphone element 11 and the microphone element 12 in advance. The target sound direction estimation unit 646 estimates the incident angle θ of the sound from the target sound direction based on the phase difference between the microphone element 11 and the microphone element 12 and the distance d between the microphone element 11 and the microphone element 12. do.

The direction determination unit 647 determines whether the direction estimated by the target sound direction estimation unit 646 is a predetermined target sound direction. If the direction estimated by the target sound direction estimation unit 646 is within a predetermined range that includes the target sound direction stored in advance, the direction determination unit 647 determines that the target sound is coming from the target sound direction. judge. On the other hand, if the direction estimated by the target sound direction estimation unit 646 is not within a predetermined range that includes the target sound direction stored in advance, the direction determination unit 647 determines that the target sound is coming from the target sound direction. It is determined that there is no. For example, the direction determination unit 647 determines whether the incident angle of the sound estimated by the target sound direction estimation unit 646 is within the range of -5 degrees to +5 degrees of the angle of the target sound direction stored in advance. You may judge. Note that a memory (not shown) stores the angle of the target sound direction in advance.

Note that in each of the above embodiments, each component may be configured with dedicated hardware, or may be realized by executing a software program suitable for each component. Each component may be realized by a program execution unit such as a CPU or a processor reading and executing a software program recorded on a recording medium such as a hard disk or a semiconductor memory.

Part or all of the functions of the device according to the embodiments of the present disclosure are typically realized as an LSI (Large Scale Integration), which is an integrated circuit. These may be integrated into one chip individually, or may be integrated into one chip including some or all of them. Further, circuit integration is not limited to LSI, and may be realized using a dedicated circuit or a general-purpose processor. An FPGA (Field Programmable Gate Array) that can be programmed after the LSI is manufactured, or a reconfigurable processor that can reconfigure the connections and settings of circuit cells inside the LSI may be used.

Further, some or all of the functions of the device according to the embodiment of the present disclosure may be realized by a processor such as a CPU executing a program.

Moreover, all the numbers used above are exemplified to specifically explain the present disclosure, and the present disclosure is not limited to the illustrated numbers.

Further, the order in which the steps shown in the above flowchart are executed is for illustrative purposes to specifically explain the present disclosure, and an order other than the above may be used as long as the same effect can be obtained. . Further, some of the above steps may be executed simultaneously (in parallel) with other steps.

The technology according to the present disclosure can improve the noise suppression performance in directional synthesis and can collect the target sound with a high S/N ratio, so the technology can collect the target sound using multiple microphone elements. Useful for sound technology.

1 Microphone array 2 Sensitivity correction unit 3, 6 Target sound detection unit 4 Sensitivity correction control unit 5 Directivity synthesis unit 11 to 1n Microphone element 31 Bandpass filter unit 32 Audio determination unit 44 Correction gain calculation unit 63 Bandpass filter unit 64, 64A Target sound direction determination section 65 Target sound determination section 201 Display 202 Air outlet 321 Level detection section 322 Noise level detection section 323 Comparison section 324 Time-frequency conversion section 325 Audio feature amount extraction section 326 Determination section 411 to 41n 1st to n Bandpass filter section 421 to 42n 1st to n level detection section 431 to 43n 1st to nth average level calculation section 641 Delay sum directivity synthesis section 642 Gradient directivity synthesis section 643 Target sound level detection section 644 Non-target sound level Detection unit 645 Level comparison and determination unit 646 Target sound direction estimation unit 647 Direction determination unit

Claims (14)

multiple microphone elements;
a sensitivity correction unit that corrects sensitivity differences between the plurality of microphone elements by multiplying output signals of the plurality of microphone elements by a gain;
a target sound detection unit that detects a speaker's voice as a target sound;
a gain control unit that controls the gain based on the detection result of the target sound detection unit;
a directional synthesis unit that emphasizes and collects the target sound coming from a predetermined direction using the output signals of the plurality of microphone elements corrected by the sensitivity correction unit;
Equipped with
The gain control unit updates the gain based on the output signals of the plurality of microphone elements when the target sound detection unit detects the voice of the speaker, and the gain control unit updates the gain based on the output signals of the plurality of microphone elements, and the target sound detection unit updates the utterance. not updating the gain if the voice of the person is not detected;
Sound collection device. The target sound detection unit includes a voice determination unit that determines whether the output signal of one of the plurality of microphone elements is the voice or non-voice other than the voice.
The sound collection device according to claim 1. The target sound detection unit includes a first extraction unit that extracts a signal in a specific band from the output signal of the one microphone element,
The voice determination unit determines whether the signal extracted by the first extraction unit is the voice or the non-voice.
The sound collection device according to claim 2. The target sound detection section includes:
a target sound direction determination unit that determines whether the target sound is coming from a predetermined target sound direction using output signals of the plurality of microphone elements;
When the target sound direction determination unit determines that the target sound is coming from the target sound direction, and the voice determination unit determines that the output signal of the one microphone element is the voice, a target sound determination unit that determines that the target sound has been detected;
including,
The sound collection device according to claim 2 or 3. The target sound detection unit includes a second extraction unit that extracts a signal in a specific band from the output signals of the plurality of microphone elements,
The target sound direction determination unit determines whether or not the target sound is coming from the target sound direction with respect to the signal extracted by the second extraction unit.
The sound collection device according to claim 4. The target sound direction determination unit includes:
a direction estimation unit that estimates a direction in which the target sound arrives using a phase difference between output signals of the plurality of microphone elements;
a direction determination unit that determines whether the direction estimated by the direction estimation unit is the predetermined target sound direction;
including,
The sound collection device according to claim 4 or 5. The target sound direction determination unit includes:
a first directivity synthesis unit that forms directivity in the direction of the target sound by emphasizing a signal in the direction of the target sound using the output signals of the plurality of microphone elements;
a second directivity synthesis unit that uses the output signals of the plurality of microphone elements to form a blind spot of sensitivity in the direction of the target sound;
Comparing the output level of the output signal from the first directional synthesis section and the output level of the output signal from the second directional synthesis section, and determining whether or not the target sound is coming from the direction of the target sound. a level comparison determination unit that determines the
including,
The sound collection device according to claim 4 or 5. The gain control section includes:
a level detection unit that detects the output level of the output signal of each of the plurality of microphone elements;
a time average level calculation unit that calculates a time average level of each output level detected by the level detection unit when the voice of the speaker is detected by the target sound detection unit;
a correction gain calculation unit that calculates a correction gain that updates the gain from the time average level calculated by the time average level calculation unit;
including,
The sound collection device according to any one of claims 1 to 7. The correction gain calculation unit calculates, based on the time average level of one microphone element predetermined among the plurality of microphone elements, the time average level of other microphone elements other than the one microphone element. calculating the correction gain of the other microphone element so that it is the same as the time average level of the one microphone element;
The sound collection device according to claim 8. The correction gain calculation unit is configured to calculate the time average level of the plurality of microphone elements such that the time average level of the plurality of microphone elements is equal to or less than the at least one microphone element based on the average value of the time average level of at least two predetermined microphone elements among the plurality of microphone elements. calculating the correction gain of the plurality of microphone elements so that it is the same as the average value of the time average level of two microphone elements;
The sound collection device according to claim 8. The gain control section includes a third extraction section that extracts a signal in a specific band from the output signal of each of the plurality of microphone elements,
The level detection section detects the output level of each signal extracted by the third extraction section.
The sound collection device according to any one of claims 8 to 10. The specific band is a band from 200Hz to 500Hz,
The sound collection device according to claim 11. The computer is
correcting sensitivity differences between the plurality of microphone elements by multiplying output signals of the plurality of microphone elements by a gain;
Detects the speaker's voice as the target sound,
controlling the gain based on the detection result of the target sound;
Emphasizing and collecting the target sound coming from a predetermined direction using the corrected output signals of the plurality of microphone elements;
In controlling the gain, when the voice of the speaker is detected, the gain is updated based on the output signals of the plurality of microphone elements, and when the voice of the speaker is not detected, the gain is updated. do not update,
Sound collection method. a sensitivity correction unit that corrects sensitivity differences between the plurality of microphone elements by multiplying output signals of the plurality of microphone elements by a gain;
a target sound detection unit that detects a speaker's voice as a target sound;
a gain control unit that controls the gain based on the detection result of the target sound detection unit;
Using the output signals of the plurality of microphone elements corrected by the sensitivity correction unit, the computer functions as a directional synthesis unit that emphasizes and collects the target sound coming from a predetermined direction;
The gain control unit updates the gain based on the output signals of the plurality of microphone elements when the target sound detection unit detects the voice of the speaker, and the gain control unit updates the gain based on the output signals of the plurality of microphone elements, and the target sound detection unit updates the utterance. not updating the gain if the voice of the person is not detected;
Sound recording program.

Images