JPH06327087A - Voice detector - Google Patents

Abstract

(57) [Abstract] [Purpose] Even if noise with a wide dynamic range is included, noise is sufficiently removed to obtain high-quality speech. [Structure] Digital first signal D including voice and noise
A digital filter unit 7 that divides the digital second signal Ds containing f and noise into a plurality of bands Za, Zb, and Zc, a plurality of noise canceling function units 5a, 5b, and 5c corresponding to each band, and each noise canceling unit. Function part 5
a, 5b, 5c digital output signals Doa, Dob,
An addition processing unit 8 for adding Doc is provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice detecting apparatus having a noise canceling function section using an adaptive filter.

[0002]

2. Description of the Related Art Generally, a voice detecting device having a noise canceling function is used for detecting voice in a noisy environment.

FIG. 5 shows a conventional voice detection device 50. The voice detection device 50 includes a first detection unit 51 including a microphone for detecting voice and noise and an analog-digital converter, a second detection unit 52 including only a microphone for detecting noise and an analog-digital converter, an adaptive filter ( An adaptive filter) 53 and a subtraction processing unit 54 are provided. Such a voice detection device 50 obtains the digital first signal Da from the first detection unit 51 and obtains the digital second signal Db from the second detection unit 52. Then, the digital second signal Db is filtered by the adaptive filter 53, and the subtraction processing unit 54 subtracts the output signal Dc of the adaptive filter 53 from the digital first signal Da, whereby the output of the subtraction processing unit 54 is obtained. A voice signal containing only the voice component with the noise component canceled is obtained.

[0004]

By the way, the principle of the above-mentioned conventional voice detecting apparatus 50 is to estimate the transfer function in the voice detecting circuit system by the detection signal from the second detecting section 52 which detects only noise. It is in. However, in general, when noise with a wide dynamic range such as white noise is included, the transfer function G with respect to the frequency f [Hz] is
The characteristic curve Mg of [dB] shows a complicated change as shown in FIG. 2 and the phase P with respect to the frequency f [Hz].
The characteristic curve Mp of [rad] also shows a complicated change as shown in FIG. For this reason, when noise with a wide dynamic range is included, the conventional voice detection device 50 has an insufficient amount of noise cancellation, and thus has a difficulty in detecting voice with high quality.

The present invention solves the problems existing in the prior art as described above. Even when noise having a wide dynamic range is included, it is possible to sufficiently remove the noise and detect high-quality speech. An object is to provide a voice detection device.

[0006]

According to the present invention, a first microphone 2 capable of detecting voice and noise, a second microphone 3 capable of detecting only noise, and digital detection signals from the respective microphones 2 and 3 are provided. One signal D
f and the analog-to-digital conversion processing unit 4 for converting into the digital second signal Ds, the digital second signal Ds is filtered by the adaptive filter, and the output signal of the adaptive filter is subtracted from the digital first signal to cancel the noise. Digital output signal Do
Noise canceling function unit 5 for outputting the digital output signal Do and the digital output signal Do for converting the analog signal Ao to the digital signal
When configuring the voice detection device 1 including the analog conversion processing unit 6, in particular, the digital first signal Df and the digital second signal Ds are respectively divided into a plurality of bands Za, Z.
b, Zc, and a plurality of noise canceling function units 5a, 5b, 5 according to each band.
c and the addition processing unit 8 for adding the digital output signals Doa, Dob, Doc of the noise canceling function units 5a, 5b, 5c.

In this case, the first microphone 2 and the second microphone 3 are integrally provided in the single casing 9 and are configured to have the same or opposite phase in each detection signal from the first microphone 2 and the second microphone 3. It is desirable that the same-sound signal component removing circuit 10 that removes the signal component is provided and configured.

[0008]

According to the voice detecting device 1 of the present invention, the first microphone 2 detects the voice including the noise, and the second microphone 3 detects only the noise, and the detection signals from the respective microphones 2 and 3 are detected. Is converted into a digital first signal Df and a digital second signal Ds by the analog-digital conversion processing unit 4, respectively.

The digital first signal Df and the digital second signal Ds are divided by the digital filter unit 7 into a plurality of bands Za, Zb, Zc, respectively. Then, a plurality of noise cancellation function units 5 corresponding to each band
Noise cancellation processing is performed by a, 5b, and 5c. That is, one noise cancellation function unit 5a
(The same applies to 5b and 5c), the digital second signal Dsa
Is filtered by the adaptive filter 11a, and the output signal Dma of the adaptive filter 11a is subtracted from the first digital signal Dfa. As a result, the noise canceling function units 5a, 5b, 5c
At the output of, the digital output signals Doa, Dob, Doc in which noise is canceled for each band are obtained. Then, the respective digital output signals Doa, Dob, Doc are added by the addition processing unit 8, and the digital output signal Do output from the addition processing unit 8 is converted into an analog signal Ao by the digital-analog conversion processing unit 7. Therefore, the analog signal Ao becomes a noise-free audio signal.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, preferred embodiments according to the present invention will be described in detail with reference to the drawings.

First, the configuration of the voice detection device 1 according to this embodiment will be described.

FIG. 1 shows a block circuit of a voice detecting device 1, 2 is a first microphone capable of detecting voice and noise, and 3 is a second microphone capable of detecting only noise. Each of the microphones 2 and 3 is integrally provided in a single casing 9 as shown in FIG. in this case,
Each of the microphones 2 and 3 uses a directional microphone, and the sound receiving unit 2 of each of the microphones 2 and 3 is used.
f and 3f are arranged so that they face each other by 180 °.
The microphones 2 and 3 may be arranged separately. Reference numeral 10 is a homophonic signal component removing circuit that removes a signal component of opposite phase (or in-phase) in each detection signal from the first microphone 2 and the second microphone 3.

Reference numeral 4 denotes an analog-to-digital conversion processing unit, which is an analog-to-digital converter 4f for converting the detection signal from the first microphone 2 output from the homophone signal component removing circuit 10 into a digital first signal Df and a homophone signal component. It is composed of an analog-digital converter 4s for converting the detection signal from the second microphone 3 output from the removing circuit 10 into the digital second signal Ds.

Reference numeral 7 designates a digital first signal Df and a digital second signal Ds in a plurality of bands Za, Zb, Zc, respectively.
It is a digital filter unit that is divided into. In the embodiment, as shown in FIG. 2, the bands Za, Zb, and Zc to be divided are set to the low band, the middle band, and the high band, respectively. Therefore, the digital filter unit 7 filters a pair of low-pass filters 12a and 13a having the same characteristic for filtering a low band, a pair of mid-band filters 12b and 13b having the same characteristic for filtering a middle band, and a high band. It is composed of a pair of high-pass filters 12c and 13c having the same characteristics. In this case, it is desirable that each of the filters 12a ... 13c uses a digital FIR filter having a linear phase characteristic, which can prevent a phase shift in each band.

Reference numeral 5 denotes a noise canceling function section, which is composed of three noise canceling function sections 5a, 5b and 5c corresponding to respective bands, that is, low band, middle band and high band. The noise canceling function unit 5a for processing the low band includes an adaptive filter 11a for filtering the digital second signal Dsa obtained from the low band filter 13a, an output signal Dma of the adaptive filter 11a, and a low band filter on the digital first signal Df side. The subtraction processing unit 14a is configured to perform a subtraction process on the output signal Dfa from 12a. On the other hand, the noise canceling function unit 5 for processing the mid range and high range
b and 5c are also configured similarly to the noise canceling function unit 5a. Note that 11b and 11c are adaptive filters, and 14b and 14c are subtraction processing units. Also, each of the adaptive filters 11 a, 11 b, 11 c is connected to the state signal generator 15.

Reference numeral 8 denotes an addition processing unit having a function of adding the digital output signals Doa, Dob, Doc output from the noise canceling function units 5a, 5b, 5c.
6 is a digital output signal Do output from the addition processing unit 8
Is a digital-to-analog conversion processing unit (digital-to-analog converter) that converts the signal into an analog signal Ao.

Next, the operation of the voice detection device 1 according to this embodiment will be described.

First, when voice is not generated, the filter coefficients of the adaptive filters 11a, 11b, 11c provided in the noise canceling function sections 5a, 5b, 5c are set (changed).

That is, in this case, the first microphone 2 detects the voice including the noise, and the second microphone 3 detects only the noise. The detection signals (analog signals) from the microphones 2 and 3 are applied to the same-sound-signal component removing circuit 10, and the opposite-phase (or same-phase) signal components in each detection signal, for example, the collision sound of an object hitting the casing 9 or the like. And so on are removed. Then, the respective detection signals output from the homophonic signal component removal circuit 10 are given to the analog-digital converters 4f and 4s, respectively, and the detection signal from the first microphone 2 is converted into the digital first signal Df, and the second The detection signal from the microphone 3 is converted into the digital second signal Ds.

On the other hand, the digital first signal Df and the digital second signal Ds are respectively filtered by the digital filter unit 7 into three bands Za, Zb and Zc of low band, middle band and high band.
Is divided into That is, from the digital first signal Df and the digital second signal Ds given to the low-pass filters 12a and 13a, the digital first signal D including only the low-frequency component
fa and the digital second signal Dsa are obtained, respectively, and the digital first signal Dfb and the digital first signal Dfb containing only the mid-range component are obtained from the digital first signal Df and the digital second signal Ds provided to the mid-range filters 12b and 13b. The two signals Dsb are respectively obtained, and further, from the digital first signal Df and the digital second signal Ds given to the high-pass filters 12c and 13c, the digital first signal Dfc and the digital second signal Dsc including only the high-frequency component are obtained. Get each.

The adaptive filter 1 provided in each of the noise canceling function sections 5a, 5b, 5c by the noise canceling function sections 5a, 5b, 5c corresponding to each band.
The filter coefficients of 1a, 11b, and 11c are changed. That is, in the low-frequency noise canceling function unit 5a,
The second digital signal Dsa is the adaptive filter 11
a, and the subtraction processing unit 14
The difference signal D obtained by subtracting the output signal Dma of the adaptive filter 11a from the digital first signal Dfa by a
The ea is obtained, and the filter coefficient of the adaptive filter 11a is changed so that the difference signal Dea becomes zero. At this time, the adaptive filter 1
A state signal Sc for changing the filter coefficient is given from 1a to the state signal generator 15. On the other hand, also in the noise canceling function units 5b and 5c, the filter coefficient is changed similarly to the noise canceling function unit 5a.

On the other hand, at the time of actual voice detection, the status signal S for fixing the filter coefficient is supplied from the status signal generator 15.
s is given to each adaptive filter 11a, 11b, 11c. In this case, the filter coefficient is fixed by the state signal Ss, but in addition, the change width (step gain) of the filter coefficient may be made smaller than the change width by the state signal Sc by the state signal Ss.
Therefore, the changing method of the filter coefficient can be changed by the state signals Sc and Ss.

By similarly picking up voice and noise from the microphones 2 and 3, the subtraction processing units 14a, 14b, and 14c receive the digital first signal Dfa,
Adaptive filters 11a and 11 from Dfb and Dfc
b, 11c output signals Doa, Dob, Doc are subtracted, and noise-cancelled digital output signal Do
It outputs a, Dab, and Dac. Then, the respective digital output signals Doa, Dob, Doc are added by the addition processing unit 8 and output as the digital output signal Do, and this digital output signal Do is converted into the analog signal Ao by the digital-analog conversion processing unit 7. , Get the desired audio signal with noise canceled.

The embodiment has been described in detail above.
The present invention is not limited to such an embodiment.
For example, voice is a concept that includes all target sounds to be detected. Further, although the embodiment exemplifies the case where the band is divided into three, it may be divided into two or four or more. On the other hand, it is not always necessary to provide the homologous signal component removing circuit. If necessary, a bandpass filter (digital FIR filter) may be inserted before the digital-analog conversion processing section. In addition, the detailed configuration, method, and the like can be arbitrarily changed without departing from the scope of the present invention.

[0025]

As described above, the voice detecting apparatus according to the present invention includes a digital filter section for dividing a digital first signal containing voice and noise and a digital second signal containing only noise into a plurality of bands, and a digital filter section. Filters the second signal with an adaptive filter, and subtracts the output signal of the adaptive filter from the digital first signal to output a digital output signal with noise canceled. Since the addition processing unit that adds the digital output signals of the cancellation function unit is provided, even if noise having a wide dynamic range is included, the noise is sufficiently removed and high-quality voice can be detected. Play.

[Brief description of drawings]

FIG. 1 is a block circuit diagram of a voice detection device according to the present invention,

FIG. 2 is a frequency-amplitude transfer function characteristic diagram of a voice detection circuit system in the voice detection device;

FIG. 3 is a frequency-to-phase transfer function characteristic diagram of a voice detection circuit system in the voice detection device,

FIG. 4 is a layout view of microphones in the voice detection device,

FIG. 5 is a block circuit diagram of a voice detection device according to a conventional technique,

[Explanation of symbols]

 1 Voice Detection Device 2 First Microphone 3 Second Microphone 4 Analog-Digital Conversion Processor 5 Noise Cancel Function 5a Noise Cancel Function 5b Noise Cancel Function 5c Noise Cancel Function 6 Digital-Analog Converter 7 Digital Filter 8 Addition processing unit 9 Casing 10 Homophone signal component removal circuit Df Digital first signal Ds Digital second signal Do Digital output signal Ao Analog signal Za band (low band) Zb band (middle band) Zc band (high band) Doa digital output Signal Dob digital output signal Doc Digital output signal

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] December 17, 1993

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be corrected] Claim 1

[Correction method] Change

[Correction content]

[Procedure Amendment 2]

[Document name to be amended] Statement

[Name of item to be corrected] 0003

[Correction method] Change

[Correction content]

FIG. 5 shows a conventional voice detection device 50. The voice detection device 50 includes a first detection unit 51 including a microphone for detecting voice and noise and an analog-digital converter, a second detection unit 52 including a microphone for detecting noise and an analog-digital converter, an adaptive filter (adaptive). The filter) 53 and the subtraction processing unit 54 are provided. Such a voice detection device 50 obtains the digital first signal Da from the first detection unit 51 and obtains the digital second signal Db from the second detection unit 52. Then, the digital second signal Db is transmitted to the adaptive filter 5
And the subtraction processing unit 5
4, the output signal Dc of the adaptive filter 53 is subtracted from the first digital signal Da, whereby an audio signal containing only the audio component in which the noise component is canceled is obtained at the output of the subtraction processing unit 54.

[Procedure 3]

[Document name to be amended] Statement

[Correction target item name] 0004

[Correction method] Change

[Correction content]

[0004]

By the way, the principle of the above-described conventional voice detecting apparatus 50 is to estimate the transfer function in the voice detecting circuit system by the detection signal from the second detecting section 52 for detecting noise. is there. However, in general, when noise with a wide dynamic range such as white noise is included, the transfer function G [d] with respect to the frequency f [Hz] is
The characteristic curve Mg of B] shows a complicated change as shown in FIG. 2 and the phase P [ra] with respect to the frequency f [Hz].
The characteristic curve Mp of d] also shows a complicated change as shown in FIG. For this reason, when noise with a wide dynamic range is included, the conventional voice detection device 50 has an insufficient amount of noise cancellation, and thus has a difficulty in detecting voice with high quality.

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0006

[Correction method] Change

[Correction content]

[0006]

According to the present invention, a first microphone 2 capable of detecting voice and noise, a second microphone 3 capable of detecting noise, and each microphone 2, 3 are provided.
The analog-to-digital conversion processing unit 4 for converting the detection signals from the respective signals into the digital first signal Df and the digital second signal Ds, and the digital second signal Ds by an adaptive filter, and from the digital first signal to the adaptive filter. In constructing the voice detection device 1 including the noise canceling function unit 5 that subtracts the output signal and outputs the digital output signal Do in which noise is canceled, in particular, the digital first signal D
f and the digital second signal Ds are respectively divided into a plurality of bands Z
a digital filter section 7 for dividing into a, Zb and Zc,
A plurality of noise canceling function units 5a, 5 corresponding to each band
b, 5c and the respective noise canceling function units 5a, 5b, 5
It is characterized by comprising an addition processing unit 8 for adding the digital output signals Doa, Dob, Doc of c.

[Procedure Amendment 5]

[Document name to be amended] Statement

[Correction target item name] 0008

[Correction method] Change

[Correction content]

[0008]

According to the voice detection device 1 of the present invention, the first microphone 2 detects the voice including the noise and the second microphone 3 detects the noise, and the detection signals from the respective microphones 2 and 3 are The analog-digital conversion processing unit 4 converts the signals into a digital first signal Df and a digital second signal Ds, respectively.

[Procedure correction 6]

[Document name to be amended] Statement

[Correction target item name] 0009

[Correction method] Change

[Correction content]

The digital first signal Df and the digital second signal Ds are divided by the digital filter unit 7 into a plurality of bands Za, Zb, Zc, respectively. Then, a plurality of noise cancellation function units 5 corresponding to each band
Noise cancellation processing is performed by a, 5b, and 5c. That is, one noise cancellation function unit 5a
(The same applies to 5b and 5c), the digital second signal Dsa
Is filtered by the adaptive filter 11a, and the output signal Dma of the adaptive filter 11a is subtracted from the first digital signal Dfa. As a result, the noise canceling function units 5a, 5b, 5c
At the output of, the digital output signals Doa, Dob, Doc in which noise is canceled for each band are obtained. Then, the digital output signals Doa, Dob, and Doc are added by the addition processing unit 8, and the digital output signal Do output from the addition processing unit 8 is converted into an analog signal Ao by the digital-analog conversion processing unit 6, for example. It Therefore, the analog signal Ao becomes a noise-free audio signal. When performing digital recording or the like, the digital output signal Do can be used directly.

[Procedure Amendment 7]

[Document name to be amended] Statement

[Correction target item name] 0012

[Correction method] Change

[Correction content]

FIG. 1 shows a block circuit of a voice detecting device 1, 2 is a first microphone capable of detecting voice and noise, and 3 is a second microphone capable of detecting noise. Each of the microphones 2 and 3 is integrally provided in a single casing 9 as shown in FIG. In this case, the microphones 2 and 3 are microphones having directivity, and the sound receiving units 2f of the microphones 2 and 3 are used.
And 3f are arranged 180 ° opposite to each other. The microphones 2 and 3 may be arranged separately. Reference numeral 10 is a homophonic signal component removing circuit that removes a signal component of opposite phase (or in-phase) in each detection signal from the first microphone 2 and the second microphone 3.

[Procedure Amendment 8]

[Document name to be amended] Statement

[Correction target item name] 0019

[Correction method] Change

[Correction content]

That is, in this case, the voice including noise is detected by the first microphone 2 and the noise is detected by the second microphone 3. The detection signals (analog signals) from the microphones 2 and 3 are applied to the same-sound-signal component removing circuit 10, and the opposite-phase (or same-phase) signal components in each detection signal, for example, the collision sound of an object hitting the casing 9 or the like. And so on are removed. And
The detection signals output from the homophonic signal component removal circuit 10 are applied to the analog-digital converters 4f and 4s, respectively, and the detection signal from the first microphone 2 is converted into the digital first signal Df and the second microphone 3 is also added. The detection signal from is converted into the digital second signal Ds.

[Procedure Amendment 9]

[Document name to be amended] Statement

[Correction target item name] 0021

[Correction method] Change

[Correction content]

The adaptive filter 1 provided in each of the noise canceling function sections 5a, 5b, 5c by the noise canceling function sections 5a, 5b, 5c corresponding to each band.
The filter coefficients of 1a, 11b, and 11c are changed. That is, in the low-frequency noise canceling function unit 5a,
The second digital signal Dsa is the adaptive filter 11
a, and the subtraction processing unit 14
The difference signal D obtained by subtracting the output signal Dma of the adaptive filter 11a from the digital first signal Dfa by a
The ea is obtained, and the filter coefficient of the adaptive filter 11a is changed so that the difference signal Dea becomes zero. At this time, the adaptive filter 1
A state signal Sc for changing the filter coefficient is given from 1a to the state signal generator 15. On the other hand, also in the noise canceling function units 5b and 5c, the filter coefficient is changed similarly to the noise canceling function unit 5a. In addition, 14b and 14c are subtraction processing units, Dmb and Dmc are output signals of the adaptive filters 11b and 11c, and De.
b and Dec indicate difference signals obtained from the subtraction processing units 14b and 14c.

[Procedure Amendment 10]

[Document name to be amended] Statement

[Name of item to be corrected] 0023

[Correction method] Change

[Correction content]

By similarly picking up voice and noise from the microphones 2 and 3, the subtraction processing units 14a, 14b, and 14c receive the digital first signal Dfa,
Adaptive filters 11a and 11 from Dfb and Dfc
b, 11c output signals Dma, Dmb, Dmc are subtracted to cancel the noise-free digital output signal Do
Output a, Dob, and Doc. Then, the respective digital output signals Doa, Dob, Doc are added by the addition processing unit 8 and output as the digital output signal Do, and this digital output signal Do is converted into the analog signal Ao by the digital-analog conversion processing unit 7. , Get the desired audio signal with noise canceled.

[Procedure Amendment 11]

[Document name to be amended] Statement

[Name of item to be corrected] 0025

[Correction method] Change

[Correction content]

[0025]

As described above, the voice detecting apparatus according to the present invention includes a digital filter unit for dividing a digital first signal containing voice and noise and a digital second signal containing noise into a plurality of bands, and a digital filter unit. Filters two signals with an adaptive filter and subtracts the output signal of the adaptive filter from the digital first signal to output a digital output signal with noise canceled. A plurality of noise cancel function units corresponding to each band and each noise cancel function. Since the addition processing unit that adds the digital output signals of the function unit is provided, even if noise with a wide dynamic range is included, the noise can be sufficiently removed and high-quality voice can be detected. Play.

[Procedure Amendment 12]

[Document name to be amended] Statement

[Name of item to be corrected] Brief description of the drawing

[Correction method] Change

[Correction content]

[Brief description of drawings]

FIG. 1 is a block circuit diagram of a voice detection device according to the present invention,

FIG. 2 is a frequency-amplitude transfer function characteristic diagram of a voice detection circuit system in the voice detection device;

FIG. 3 is a frequency-to-phase transfer function characteristic diagram of a voice detection circuit system in the voice detection device,

FIG. 4 is a layout view of microphones in the voice detection device,

FIG. 5 is a block circuit diagram of a voice detection device according to a conventional technique,

[Description of Reference Signs] 1 voice detection device 2 first microphone 3 second microphone 4 analog-digital conversion processing unit 5 noise canceling function unit 5a noise canceling function unit 5b noise canceling function unit 5c noise canceling function unit 7 digital filter unit 8 addition Processor 9 Casing 10 Homophone signal component removal circuit Df Digital first signal Ds Digital second signal Do Digital output signal Za band (low band) Zb band (middle band) Zc band (high band) Doa digital output signal Dob digital output signal Doc Digital output signal

Claims (3)

[Claims] 1. A first microphone capable of detecting voice and noise, a second microphone capable of detecting only noise, and an analog for converting a detection signal from each microphone into a digital first signal and a digital second signal, respectively. A digital conversion processing unit, a noise canceling function unit that filters the digital second signal with an adaptive filter, and subtracts the output signal of the adaptive filter from the digital first signal to output a digital output signal with noise canceled, and a digital output In a voice detection device including a digital-analog conversion processing unit for converting a signal into an analog signal, a digital filter unit for dividing each of the digital first signal and the digital second signal into a plurality of bands, and a plurality of units for each band. Noise canceling A voice detecting device, comprising: a function unit and an addition processing unit for adding digital output signals of the noise canceling units. 2. The voice detecting device according to claim 1, wherein the first microphone and the second microphone are integrally provided in a single casing. 3. The voice detecting device according to claim 2, further comprising an homologous signal component removing circuit for eliminating in-phase or anti-phase signal components in each detection signal from the first microphone and the second microphone.