JP2013192087A - Noise suppression device, microphone array device, noise suppression method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent non-stationary noise from being emphasized, when directional sound reception processing is performed under an environment where high level non-stationary noise, such as wind blow noise, is generated.SOLUTION: A correlation coefficient calculation unit 12a calculates a correlation coefficient r(t) of input signals Lin(t) and Rin(t), input from a microphone array 11. An analysis unit 12b and a reduction unit 12c output the signal Lin'(t) and Rin'(t) produced by performing non-stationary noise reduction processing of an input signal on the basis of the r(t). A synchronous subtraction coefficient calculation unit 14 calculates a synchronous subtraction coefficient α(t) which decreases as the absolute value |r(t)| of the correlation coefficient decreases. A delay unit 16a outputs a signal Rin'(t-1) by delaying a signal Rin'(t) by one sample. A multiplication unit 16b outputs a signal α(t)Rin'(t-1) produced by multiplying the signal Rin'(t-1) by α(t). A subtraction unit 16c outputs a signal produced by subtracting the signal α(t)Rin'(t-1) from the signal Lin'(t).

Description

  The present invention relates to a noise suppression device, a microphone array device, a noise suppression method, and a program.

  Conventionally, as shown in FIG. 15, a microphone array that performs directional sound reception processing by synchronous subtraction or synchronous addition of outputs from two microphones is generally used. For example, a microphone array is installed between the driver's seat and the passenger's seat in the car, and the sound coming from that direction is suppressed as noise by synchronous subtraction according to the passenger's seat direction. Can have sex.

  In addition, when the wind blows from the air conditioner or the like on the microphone, the diaphragm moves so much that a large level of unsteady noise occurs. Techniques for suppressing such a large level of non-stationary noise have been proposed. For example, in the example illustrated in FIG. 16, the correlation analysis unit 1000 calculates a correlation between input signals input from a plurality of microphones. When the correlation calculated by the correlation analysis unit 1000 is small, the wind noise analysis unit 1002 determines that noise due to wind contact has occurred, and outputs a reduction gain to the wind noise reduction unit 1004. The wind noise reduction unit 1004 outputs the input signal with the level reduced according to the reduction gain (see, for example, Patent Documents 1 to 3).

Japanese Patent Laid-Open No. 10-23590 Japanese Patent No. 2995959 Japanese Patent Laid-Open No. 2008-263483

  When directional sound reception processing is performed in an environment where a large level of non-stationary noise such as wind noise occurs, this non-stationary noise is emphasized by synchronous subtraction or synchronous addition for directional sound reception processing. There is a problem. This is because, in the case of a large level of non-stationary noise such as wind noise, the correlation between the input signals from each of the plurality of microphones is small, so that the power obtained as a result of synchronous subtraction is as follows. This is because it becomes a value equivalent to the sum of the powers of. Similarly, in the synchronous subtraction, the noise power added by shifting the phase becomes a value equivalent to the sum of the powers of the input signals.

"(Lin (t) -Rin (t-1)) power" ≒ "Lin (t) power" + "Rin (t) power"
Here, Lin (t) and Rin (t) are input signals, and Rin (t−1) is a signal obtained by delaying the input signal Rin (t) by one sample.

  Further, in the case where the directional sound reception and the unsteady noise reduction are simultaneously realized, it is conceivable to combine the wind noise reduction process and the directional sound reception process as shown in FIG. However, even in this case, there is a problem that the residual wind noise is emphasized by synchronous subtraction or synchronous addition for directivity reception.

  FIG. 18 shows an example in which wind noise is enhanced by synchronous subtraction. It shows that wind noise is generated at the location where the amplitude is protruding. In the example of FIG. 18, it can be seen that wind noise is emphasized after synchronization subtraction compared to before synchronization subtraction. Here, the problem of an increase in noise level of about 6 dB occurs.

  The disclosed technique is intended to suppress the enhancement of unsteady noise when performing directional sound reception processing in an environment in which a large level of unsteady noise such as wind noise is generated. .

  In the disclosed technique, a correlation analysis unit analyzes a correlation between a plurality of input signals input from each of a plurality of microphones. The directional sound receiving unit performs directional sound reception processing on the plurality of input signals so that the degree of synchronization subtraction or synchronization addition decreases as the correlation analyzed by the correlation analysis unit decreases. .

  In addition, the disclosed technology can include a microphone array in which a plurality of microphones are arranged at predetermined intervals.

  The disclosed technique has an effect that it is possible to suppress the enhancement of the non-stationary noise in the case where the directional sound reception process is performed in an environment where a large level of non-stationary noise occurs.

1 is a functional block diagram of a microphone array apparatus according to a first embodiment. It is a figure for demonstrating the process of the synchronous subtraction coefficient calculation part of 1st Embodiment. It is a flowchart which shows the content of the noise suppression process in the microphone array apparatus which concerns on 1st Embodiment. It is a figure which shows the synchronous subtraction result in the microphone array apparatus which concerns on this Embodiment compared with the prior art. It is a functional block diagram of the microphone array apparatus which concerns on 2nd Embodiment. It is a figure for demonstrating the process of the synchronous subtraction coefficient calculation part of 2nd Embodiment. It is a flowchart which shows the content of the noise suppression process in the microphone array apparatus which concerns on 2nd Embodiment. It is a functional block diagram of the microphone array apparatus which concerns on 3rd Embodiment. It is a figure for demonstrating the process of the synchronous subtraction coefficient calculation part of 3rd Embodiment. It is a flowchart which shows the content of the noise suppression process in the microphone array apparatus which concerns on 3rd Embodiment. It is a figure which shows parameter adjustment according to a microphone space | interval. It is a functional block diagram of the microphone array apparatus which concerns on 5th Embodiment. It is a figure which shows parameter adjustment based on the relative value of output power. It is a schematic block diagram of the computer which functions as a microphone array apparatus concerning a 6th embodiment. It is a block diagram which shows the directivity sound reception process by the synchronous subtraction in a prior art. It is a block diagram which shows the wind noise reduction process based on the correlation between the input signals in a prior art. It is a block diagram which combined the directivity sound reception process and wind noise reduction process in a prior art. It is a figure which shows the problem of the directivity sound reception process of a prior art.

  Hereinafter, an example of an embodiment of the disclosed technology will be described in detail with reference to the drawings.

  <First Embodiment>

  FIG. 1 shows a microphone array apparatus 1 according to the first embodiment. The microphone array device 1 includes a microphone array 11 in which a plurality of microphones are arranged at a predetermined interval, and a noise suppression device 10.

  The microphone array 11 includes at least two microphones. Here, an example will be described in which two microphones, an L channel microphone 11a and an R channel microphone 11b, are included.

  The microphones 11a and 11b collect surrounding sounds and convert the collected sounds into analog signals. Also, the analog signal is converted into a digital signal sampled at a predetermined sampling frequency Fs, and output as input signals Lin (t) and Rin (t), respectively. Note that t is a sampling number.

  The noise suppression device 10 can be represented by including a non-stationary noise reduction unit 12, a synchronous subtraction coefficient calculation unit 14, and a directivity sound reception unit 16. Each functional unit of the noise suppression device 10 can be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC (Application Specific Integrated Circuit). The noise suppression apparatus 10 includes a DSP (Digital Signal Processor) to which a digital signal output from the microphone array 11 is input, and a digital / analog converter that converts the digital signal output from the DSP into an analog signal. You can also

  The non-stationary noise reduction unit 12 can further be represented by including a correlation coefficient calculation unit 12a, an analysis unit 12b, and a reduction unit 12c. The correlation coefficient calculation unit 12a is an example of a correlation analysis unit of the disclosed technology, and the analysis unit 12b and the reduction unit 12c are examples of the reduction unit of the disclosed technology.

  The correlation coefficient calculator 12a calculates a correlation coefficient r (t) between the input signals Lin (t) and Rin (t) input from the microphones 11a and 11b by the following equation (1). Note that the variable i in the equation (1) is a natural number of 0 to I, and indicates that the correlation coefficient is calculated using the input signals in each sample from (t−I) to t.

r (t) = Σ _{i = 0} Lin (ti) Rin (ti) / (Σ _{i = 0} Lin (ti) ² Σ _{i = 0} Rin (ti) ² ) ^1/2 (1)

  When a large level of non-stationary noise such as wind noise occurs, the correlation between the two input signals Lin (t) and Rin (t) becomes small. For example, when a driver's voice is collected by a microphone array mounted in a vehicle, the waveform of the input signal from each microphone from which the driver's voice is collected is similar, and the correlation between the input signals increases. On the other hand, when wind from an air conditioner hits the microphone, the waveform of the input signal from each microphone is different, and the correlation between the input signals is small. Therefore, the absolute value | r (t) | of the correlation coefficient is an index indicating the certainty that non-stationary noise is generated. That is, the smaller the absolute value | r (t) | of the correlation coefficient, the higher the probability that non-stationary noise is generated.

  Based on the absolute value | r (t) | of the correlation coefficient calculated by the correlation coefficient calculation unit 12a, the analysis unit 12b analyzes whether or not a large level of non-stationary noise is generated in the input signal. A reduction gain corresponding to the absolute value | r (t) | of the correlation coefficient is output. The reduction gain is a value that increases as the absolute value | r (t) | of the correlation coefficient increases and decreases as the absolute value | r (t) | of the correlation coefficient decreases.

  The reduction unit 12c uses the reduction gain output from the analysis unit 12b to obtain signals Lin ′ (t) and Rin ′ (t) obtained by reducing the levels of the input signals Lin (t) and Rin (t). Output. Thus, as the probability that a large level of non-stationary noise is generated is higher, the levels of the input signals Lin (t) and Rin (t) are reduced, and the non-stationary noise is reduced.

  The synchronous subtraction coefficient calculation unit 14 calculates a synchronous subtraction coefficient in a directivity sound receiving unit 16 to be described later. When a large level of non-stationary noise such as wind noise is generated, the non-stationary noise is emphasized by synchronous subtraction for directional sound reception processing. Therefore, when there is a high probability that a large level of non-stationary noise has occurred, the degree of the synchronous subtraction is reduced to suppress the enhancement of the non-stationary noise due to the synchronous subtraction. The synchronous subtraction coefficient calculation unit 14 is an example of a coefficient calculation unit of the disclosed technique.

  The synchronous subtraction coefficient calculator 14 calculates a synchronous subtraction coefficient α (t) that indicates the degree of the synchronous subtraction. For example, as shown in FIG. 2, the synchronous subtraction coefficient α (t) is determined so as to increase as the absolute value | r (t) | of the correlation coefficient increases and decrease as | r (t) | Keep it. That is, the synchronous subtraction coefficient α (t) becomes smaller as the probability that a large level of non-stationary noise is generated is higher. Note that 0 ≦ | r (t) | ≦ 1 and 0 ≦ α (t) ≦ 1. The relationship between the synchronous subtraction coefficient α (t) and the absolute value | r (t) | of the correlation coefficient as shown in FIG. 2 is stored in a predetermined storage area in the form of a table or a function, for example.

  The synchronous subtraction coefficient calculation unit 14 refers to the relationship between the synchronous subtraction coefficient α (t) stored in a predetermined storage area and the absolute value | r (t) | of the correlation coefficient, from the correlation coefficient calculation unit 12a. A synchronous subtraction coefficient α (t) corresponding to the absolute value | r (t) | of the input correlation coefficient is calculated. The calculated synchronous subtraction coefficient α (t) is output to the directional sound receiving unit 16.

  The directivity sound receiving unit 16 can further be represented by including a delay unit 16a, a multiplication unit 16b, and a subtraction unit 16c.

  The delay unit 16a delays the signal Rin ′ (t) input from the non-stationary noise reduction unit 12 by a delay amount d for synchronization at the time of synchronous subtraction determined based on the interval between the microphones 11a and 11b. Signal Rin ′ (t−d) is output. For example, when the interval between the microphones 11a and 11b is the sound speed c / sampling frequency Fs, the signal Rin ′ (t) is delayed by one sample, thereby synchronizing synchronization subtraction. In this case, the delay unit 16a can be configured as a one-sample delay device. In the first embodiment, a case where the delay unit 16a is a one-sample delay device will be described as an example. That is, the signal Rin ′ (t−1) is output from the delay unit 16a.

  The multiplication unit 16b multiplies the signal Rin ′ (t−1) output from the delay unit 16a by the synchronous subtraction coefficient α (t) calculated by the synchronous subtraction coefficient calculation unit 14 to obtain a signal α (t) Rin ′ ( t-1) is output.

  The subtraction unit 16c outputs a signal obtained by subtracting the signal α (t) Rin ′ (t−1) output from the multiplication unit 16b from the signal Lin ′ (t) input from the unsteady noise reduction unit 12. Output as (t). The output signal out (t) is expressed by the following equation (2).

    out (t) = Lin '(t) -α (t) Rin' (t-1) (2)

  Next, the operation of the microphone array device 1 of the first embodiment will be described. When ambient sounds are collected by the microphone array 11 and input signals Lin (t) and Rin (t) are input to the noise suppression device 10 from the microphones 11a and 11b, respectively, in the noise suppression device 10, FIG. The noise suppression process shown is executed.

  First, in step 100, the correlation coefficient calculation unit 12a uses the equation (1) to calculate the correlation coefficient r (t) between the input signals Lin (t) and Rin (t) input from each of the microphones 11a and 11b. Calculate

  Next, in step 102, the analysis unit 12b determines whether or not a large level of unsteady noise is generated in the input signal based on the absolute value | r (t) | of the correlation coefficient calculated in step 100. And a reduction gain corresponding to the absolute value | r (t) | of the correlation coefficient is output. Then, the reduction unit 12c uses the reduction gain output from the analysis unit 12b to reduce the level of the input signals Lin (t) and Rin (t), and the signals Lin ′ (t) and Rin ′ (t ) Is output.

  Next, in step 104, the synchronous subtraction coefficient calculator 14 refers to the relationship between the synchronous subtraction coefficient α (t) stored in a predetermined storage area and the absolute value | r (t) | of the correlation coefficient. Then, the synchronous subtraction coefficient α (t) corresponding to the absolute value | r (t) | of the correlation coefficient calculated in step 100 is calculated.

  Next, in step 106, the delay unit 16a outputs a signal Rin '(t-1) obtained by delaying the signal Rin' (t) by one sample. Then, the multiplication unit 16b multiplies the signal Rin ′ (t−1) output from the delay unit 16a by the synchronous subtraction coefficient α (t) calculated in Step 104 above to obtain a signal α (t) Rin ′ (t -1) is output. Then, the subtracting unit 16c performs the signal α (t) Rin ′ (t−1) output from the multiplying unit 16b from the signal Lin ′ (t) in which the non-stationary noise is reduced in step 102 according to the equation (2). ) Is output as an output signal out (t), and the process ends.

  As described above, according to the microphone array apparatus of the first embodiment, the larger the absolute value of the correlation coefficient between the input signals indicating the probability of occurrence of a large level of non-stationary noise, the larger the synchronization. The subtraction coefficient is calculated, and this synchronous subtraction coefficient is used in the synchronous subtraction in the directivity receiving process. For this reason, when the probability that non-stationary noise is generated is high, it is possible to suppress enhancement of non-stationary noise by synchronous subtraction by reducing the degree of synchronous subtraction.

  FIG. 4 shows the result of synchronous subtraction of fixed coefficients in the environment with wind noise (same as the right diagram of FIG. 18) and the result of synchronous subtraction according to the present embodiment. In the example of FIG. 4, according to the present embodiment, it was found that the enhancement of wind noise can be suppressed by about 6 dB.

  <Second Embodiment>

  Next, a second embodiment will be described. FIG. 5 shows a microphone array apparatus 2 according to the second embodiment. Note that the same portions as those of the microphone array device 1 of the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

  The microphone array device 2 according to the second embodiment includes a microphone array 11 and a noise suppression device 210. The noise suppression apparatus 210 can be represented including an unsteady noise reduction unit 212, a synchronous subtraction coefficient calculation unit 214, and a directivity sound reception unit 216.

  The non-stationary noise reduction unit 212 may further include a correlation coefficient calculation unit 212a, a sound source direction detection unit 12d, an analysis unit 12b, and a reduction unit 12c.

  The correlation coefficient calculation unit 212a calculates a correlation coefficient r (t, d) between the input signals Lin (t) and Rin (t) input from the microphones 11a and 11b by the following equation (3). .

r (t, d) = Σ _{i = 0} Lin (ti) Rin (tid) / (Σ _{i = 0} Lin (ti) ² Σ _{i = 0} Rin (tid) ² ) ^1/2 (3)

Here, d is a delay amount. For example, when the distance between the microphones 11a and 11b is the sound velocity c / sampling frequency Fs, d = −1, 0, 1. The correlation coefficient calculation unit 212a calculates the correlation coefficient when the delay amount d is each value by the expression (3), and the value | r (t, d) | when the absolute value | r (t, d) | (T, d _max ) | is output.

The sound source direction detection unit 12d has the _maximum value of d _max in the absolute value | r (t, d _max ) | of the correlation coefficient output from the correlation coefficient calculation unit 212a, that is, the absolute value of the correlation coefficient of the equation (3) The sound source direction is detected based on the delay amount d _max when When d _max = −1, the L channel microphone 11a side is the sound source direction. When d _max = 0, an intermediate direction between the L-channel microphone 11a and the R-channel microphone 11b is the sound source direction. In the case of d _max = 1, the R channel microphone 11b side is the sound source direction.

  Further, the sound source direction detection unit 12d obtains delay amounts dL and dR for synchronization according to the detected sound source direction for each of the signals Lin ′ (t) and Rin ′ (t) as follows.

When d _max ≧ 0: dL = 0, dR = d _max
When d _max <0: dL = | d _max |, dR = 0

Similar to the relationship between the synchronous subtraction coefficient α (t) and the absolute value of the correlation coefficient | r (t) | in the first embodiment, the synchronous subtraction coefficient calculation unit 214, for example, as shown in FIG. A predetermined relationship between the synchronous subtraction coefficient α (t) and the absolute value | r (t, d _max ) | of the correlation coefficient is used.

The synchronous subtraction coefficient calculation unit 214 refers to the relationship between the synchronous subtraction coefficient α (t) stored in a predetermined storage area and the absolute value | r (t, d _max ) | of the correlation coefficient. Then, a synchronous subtraction coefficient α (t) corresponding to the absolute value | r (t, d _max ) | of the correlation coefficient input from the correlation coefficient calculation unit 212a is calculated. The calculated synchronous subtraction coefficient α (t) is output to the directional sound receiving unit 216.

  The directivity sound receiving unit 216 can further be represented by including a delay unit 216aL, a delay unit 216aR, a multiplication unit 216b, and a subtraction unit 216c.

  The delay unit 216aL outputs a signal Lin ′ (t−dL) obtained by delaying the signal Lin ′ (t) input from the non-stationary noise reduction unit 212 by the delay amount dL obtained by the sound source direction detection unit 12d. . Similarly, the delay unit 216aR is a signal Rin ′ (t−dR) obtained by delaying the signal Rin ′ (t) input from the non-stationary noise reduction unit 212 by the delay amount dR obtained by the sound source direction detection unit 12d. Is output.

  Similarly to the first embodiment, the multiplication unit 216b adds the synchronous subtraction coefficient α (t) calculated by the synchronous subtraction coefficient calculation unit 214 to the signal Rin ′ (t−dR) input from the delay unit 216aR. The multiplied signal α (t) Rin ′ (t−dR) is output.

  The subtraction unit 216c subtracts the signal α (t) Rin ′ (t−dR) input from the multiplication unit 216b from the signal Lin ′ (t−dL) input from the delay unit 216aL as an output signal out ( t). The output signal out (t) is expressed by the following equation (4).

    out (t) = Lin '(t-dL) -α (t) Rin' (t-dR) (4)

  Next, the operation of the microphone array apparatus 2 of the second embodiment will be described. When ambient sounds are collected by the microphone array 11 and the input signals Lin (t) and Rin (t) are input from the microphones 11a and 11b to the noise suppression device 210, the noise suppression device 210 shows in FIG. The noise suppression process shown is executed. In addition, about the process similar to the noise suppression process in 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

First, in step 200, the correlation coefficient calculation unit 212a uses the equation (3) to calculate a correlation coefficient for each delay amount between the input signals Lin (t) and Rin (t) input from the microphones 11a and 11b. Calculate r (t, d). Then, a value | r (t, d _max ) | when the absolute value becomes maximum is output.

Next, in step 202, the sound source direction detection unit 12d detects the sound source direction based on d _max in the absolute value | r (t, d _max ) | of the correlation coefficient output in step 200. Then, for each of the signals Lin ′ (t) and Rin ′ (t) output in the next step, delay amounts dL and dR for synchronization according to the detected sound source direction are obtained.

Next, in step 102, the analysis unit 12b and the reduction unit 12c cause the signal Lin from which unsteady noise is reduced based on the absolute value | r (t, d _max ) | of the correlation coefficient calculated in step 200 above. '(T) and Rin' (t) are output.

Next, in step 204, the synchronous subtraction coefficient calculation unit 214 determines the relationship between the synchronous subtraction coefficient α (t) stored in the predetermined storage area and the absolute value | r (t, d _max ) | of the correlation coefficient. refer. Then, the synchronous subtraction coefficient α (t) corresponding to the absolute value | r (t, d _max ) | of the correlation coefficient calculated in step 200 is calculated.

  Next, in Step 206, the delay units 216aL and 216aR cause the signals Lin ′ (t) and Rin ′ (t) to be delayed by the delay amounts dL and dR obtained in Step 202 and the signal Lin ′ (t− dL) and Rin ′ (t−dR) are output. Then, the multiplication unit 16b multiplies the signal Rin ′ (t−dR) input from the delay unit 216aR by the synchronous subtraction coefficient α (t) calculated in step 204 to obtain a signal α (t) Rin ′ (t -DR) is output. Then, the subtracting unit 216c calculates the signal α (t) Rin ′ (t−dR) input from the multiplying unit 216b from the signal Lin ′ (t−dL) input from the delay unit 216aL according to the equation (4). The subtracted signal is output as an output signal out (t), and the process is terminated.

  As described above, according to the microphone array device 2 of the second embodiment, the sound source direction is detected from the delay amount when the absolute value of the correlation coefficient of the input signal obtained for each delay amount is maximized. Thus, the delay amount at the time of synchronous subtraction is obtained. Thereby, the directivity of directivity sound reception can be controlled. Similarly to the first embodiment, a synchronous subtraction coefficient that is smaller as the absolute value of the correlation coefficient between the input signals indicating the probability of occurrence of a large level of unsteady noise is smaller is calculated. This synchronous subtraction coefficient is used in the synchronous subtraction in the sound receiving process. For this reason, when the probability that non-stationary noise is generated is high, it is possible to suppress enhancement of non-stationary noise by synchronous subtraction by reducing the degree of synchronous subtraction.

  In addition, when there are a plurality of sound source directions, the above processing may be performed in parallel according to each sound source direction.

  <Third Embodiment>

  Next, a third embodiment will be described. FIG. 8 shows a microphone array apparatus 3 according to the third embodiment. Note that the same portions as those of the microphone array device 1 of the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

  The microphone array device 3 according to the third embodiment includes a microphone array 11 and a noise suppression device 310. The noise suppression device 310 can be represented including an FFT unit 18L, 18R, an unsteady noise reduction unit 312, a synchronous subtraction coefficient calculation unit 314, a directivity sound reception unit 316, and an IFFT unit 20. Note that the FFT units 18L and 18R are examples of the frequency domain conversion unit of the disclosed technology.

  The FFT units 18L and 18R convert the input signals Lin (t) and Rin (t), which are time domain signals, into frequency domain signals. Specifically, the Hanning window overlapped at a predetermined ratio (for example, 50%) is performed on each of the input signals Lin (t) and Rin (t), and the frequency domain is calculated by FFT (Fast Fourier Transform). Convert to signals LIN (k, f) and RIN (k, f). Here, k indicates a frame number, and f indicates a frequency number.

  The non-stationary noise reduction unit 312 can further be represented by including a correlation coefficient calculation unit 312a, an analysis unit 12b, and a reduction unit 12c.

  The correlation coefficient calculation unit 312a calculates the correlation coefficient r (k, f) between the input signal LIN (k, f) and RIN (k, f) converted into the frequency domain signal by the following equation (5). calculate.

r (k, f) = Σ _{i = 0} LIN (ki, f) RIN (ki, f) / (Σ _{i = 0} LIN (ki, f) ² Σ _{i = 0} RIN (ki, f) ² ) ^{1 / 2} (5)

  Based on the absolute value | r (k, f) | of the correlation coefficient calculated by the correlation coefficient calculation unit 312a, the analysis unit 12b and the reduction unit 12c reduce the signal LIN ′ (k, f ) And RIN ′ (k, f).

  Similar to the relationship between the synchronous subtraction coefficient α (t) and the absolute value | r (t) | of the correlation coefficient in the first embodiment, the synchronous subtraction coefficient calculation unit 314, for example, as shown in FIG. A predetermined relationship between the synchronous subtraction coefficient α (k, f) and the absolute value | r (k, f) | of the correlation coefficient is used.

  The synchronous subtraction coefficient calculation unit 314 refers to the relationship between the synchronous subtraction coefficient α (k, f) stored in a predetermined storage area and the absolute value | r (k, f) | of the correlation coefficient. Then, the synchronous subtraction coefficient α (k, f) corresponding to the absolute value | r (k, f) | of the correlation coefficient input from the correlation coefficient calculation unit 312a is calculated. The calculated synchronous subtraction coefficient α (k, f) is output to the directivity sound receiving unit 316.

  The directivity sound receiving unit 316 can further be represented by including a delay unit 316a, a multiplication unit 316b, and a subtraction unit 316c.

  The delay unit 316a receives the signal RIN ′ (k, f) input from the non-stationary noise reduction unit 312 by a delay amount d for synchronization at the time of synchronous subtraction determined based on the interval between the microphones 11a and 11b. Output delayed signal. For example, when the interval between the microphones 11a and 11b is the sound velocity c / sampling frequency Fs, the synchronization subtraction can be synchronized by delaying the signal RIN ′ (k, f) by one sample. In this case, the delay unit 316a can be configured as a one-sample delay unit, and the signal exp (−2πf / Fs) RIN ′ (k, f) is output from the delay unit 316a. Here, Fs is a sampling frequency, and exp (−2πf / Fs) is a phase spectrum shift coefficient in the frequency domain corresponding to one sample delay in the time domain.

  The multiplier 316b multiplies the signal exp (−2πf / Fs) RIN ′ (k, f) output from the delay unit 316a by the synchronous subtraction coefficient α (k, f) calculated by the synchronous subtraction coefficient calculator 314. The signal α (k, f) exp (−2πf / Fs) RIN ′ (k, f) is output.

  The subtractor 316c receives the signal α (k, f) exp (−2πf / Fs) RIN ′ (k) input from the multiplier 316b from the signal LIN ′ (k, f) input from the unsteady noise reduction unit 312. , F) is output as an output signal OUT (k, f). The output signal OUT (k, f) is expressed by the following equation (6).

    OUT (k, f) = LIN ′ (k, f) −α (k, f) exp (−2πf / Fs) RIN ′ (k, f) (6)

  The IFFT unit 20 converts the output signal OUT (k, f), which is a frequency domain signal input from the directional sound receiving unit 316, into a time domain signal by IFFT (Inverse FFT), and a predetermined ratio (for example, 50%), an output signal out (t) is output.

  Next, the operation of the microphone array device 3 according to the third embodiment will be described. When ambient sounds are collected by the microphone array 11 and the input signals Lin (t) and Rin (t) are input from the microphones 11a and 11b to the noise suppression device 310, the noise suppression device 310 shows in FIG. The noise suppression process shown is executed. In addition, about the process similar to the noise suppression process in 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  First, in step 300, the FFT units 18L and 18R perform Hanning windowing that overlaps each of the input signals Lin (t) and Rin (t) at a predetermined ratio, and the frequency domain signal LIN ( k, f) and RIN (k, f) for conversion.

  Next, in step 302, the correlation coefficient calculation unit 312a uses the equation (5) to calculate the correlation coefficient r between the input signal Lin (k, f) and Rin (k, f) converted into the frequency domain signal. Calculate (k, f).

  Next, in step 102, the analysis unit 12b and the reduction unit 12c have the signal LIN ′ with reduced unsteady noise based on the absolute value | r (k, f) | of the correlation coefficient calculated in step 300. (K, f) and RIN ′ (k, f) are output.

  Next, in step 304, the synchronous subtraction coefficient calculation unit 314 has a relationship between the synchronous subtraction coefficient α (k, f) stored in a predetermined storage area and the absolute value | r (k, f) | of the correlation coefficient. Refer to Then, the synchronous subtraction coefficient α (k, f) corresponding to the absolute value | r (k, f) | of the correlation coefficient calculated in step 300 is calculated.

  Next, in step 306, the delay unit 316a outputs a signal exp (−2πf / Fs) RIN ′ (k, f) obtained by delaying RIN ′ (k, f) by one sample. Then, the multiplication unit 316b multiplies the signal exp (−2πf / Fs) RIN ′ (k, f) output from the delay unit 316a by the synchronous subtraction coefficient α (k, f) calculated in step 304 above. The signal α (k, f) exp (−2πf / Fs) RIN ′ (k, f) is output. Then, the subtracting unit 316c, from the signal LIN ′ (k, f) input from the unsteady noise reduction unit 312 according to the equation (6), the signal α (k, f) exp (−) input from the multiplying unit 316b. A signal obtained by subtracting 2πf / Fs) RIN ′ (k, f) is output as an output signal OUT (k, f).

  Next, in step 308, the IFFT unit 20 converts the output signal OUT (k, f), which is the frequency domain signal output in step 306, into a time domain signal by IFFT, and overlaps at a predetermined rate. The added output signal out (t) is output.

  As described above, according to the microphone array device 3 of the third embodiment, by converting the input signal into a frequency domain signal, the synchronous subtraction coefficient is calculated for each frequency component and the synchronous subtraction process is performed. It can be carried out. Similarly to the first embodiment, a synchronous subtraction coefficient that is smaller as the absolute value of the correlation coefficient between the input signals indicating the probability of occurrence of a large level of unsteady noise is smaller is calculated. This synchronous subtraction coefficient is used in the synchronous subtraction in the sound receiving process. For this reason, when the probability that non-stationary noise is generated is high, it is possible to suppress enhancement of non-stationary noise by synchronous subtraction by reducing the degree of synchronous subtraction.

  <Fourth embodiment>

  Next, a fourth embodiment will be described. Since the configuration of the microphone array apparatus 4 according to the fourth embodiment is the same as that of the microphone array apparatus 1 according to the first embodiment, the description thereof is omitted.

  In the microphone array device 4 according to the fourth embodiment, the parameters for calculating the synchronous subtraction coefficient α by the synchronous subtraction coefficient calculation unit 14 are determined according to the interval between the microphones 11a and 11b.

  When the distance between the microphones 11a and 11b is short, the correlation between the input signals tends to be high even when a large level of unsteady noise such as wind noise is generated. In consideration of such a case, when the microphone interval is short, a parameter for calculating the synchronization subtraction coefficient α is determined so that the synchronization subtraction coefficient α is increased when the correlation is truly high.

  For example, the parameters γ0 and γ1 in the relationship between the synchronous subtraction coefficient α and the absolute value | r | of the correlation coefficient as shown in FIGS. 2, 6, and 9 are adjusted as shown in FIG. 11, for example. . In the example of FIG. 11, the horizontal axis represents the relative value of the microphone interval with “sound speed c / sampling frequency Fs” as a reference, and the vertical axis represents γ0 and γ1. Even when a large level of non-stationary noise is generated, the absolute value | r | of the correlation coefficient becomes closer to 1.0 as the microphone interval is shorter. Therefore, as shown in FIG. 11, the values of the parameters γ0 and γ1 Is set to a value close to 1.0.

  As described above, according to the microphone array device 4 of the fourth embodiment, an appropriate synchronous subtraction coefficient is calculated in consideration of the case where the correlation coefficient of the input signal increases due to the short microphone interval. Therefore, it is possible to appropriately suppress the enhancement of non-stationary noise due to synchronous subtraction.

  <Fifth embodiment>

  Next, a fifth embodiment will be described. FIG. 12 shows a microphone array apparatus 5 according to the fifth embodiment. Note that the same portions as those of the microphone array device 1 of the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

  The microphone array device 5 according to the fifth embodiment includes a microphone array 11 and a noise suppression device 510. The noise suppression device 510 can be represented including an unsteady noise reduction unit 12, a synchronous subtraction coefficient calculation unit 514, a directivity sound reception unit 16, and a parameter update unit 22. The parameter update unit 22 is an example of an update unit of the disclosed technology.

  The parameter update unit 22 is based on the relative value of the output power represented by the ratio of the power of the output out (t) to the power of the input Lin ′ (t) to the directivity receiving unit 16, and the synchronous subtraction coefficient calculation unit In 514, a parameter for calculating the synchronous subtraction coefficient α is updated.

  The relative value of the output power in accordance with the synchronous subtraction coefficient α varies depending on the wind contact condition and the traveling condition of the vehicle in which the microphone array device 5 is mounted in the vehicle. For example, as shown in FIG. 13, when the horizontal axis represents the synchronous subtraction coefficient α and the vertical axis represents the relative value of the output power, the α value αopt (r (t )) Changes. In addition, that the relative value of output power is small shows that the noise suppression effect by a directional sound reception process is high.

  The parameter updating unit 22 calculates the value of αopt (r (t)) at different correlation coefficients r (t), and sets a plurality of different correlation coefficients r (t) and αopt (r (t)). The relationship between the correlation coefficient r (t) and αopt (r (t)) is obtained by linear interpolation based on the above. Based on this relationship, the values of the parameters α0 and α1 corresponding to the parameters γ0 and γ1 are acquired, and the parameters α0 and α1 set in the synchronous subtraction coefficient calculation unit 514 are updated.

  As described above, according to the microphone array device 5 of the fifth embodiment, the parameters for calculating the synchronous subtraction coefficient are updated so that the noise suppression effect is enhanced. It is possible to suppress the enhancement of noise and further enhance the noise suppression effect.

  <Sixth Embodiment>

  Next, a sixth embodiment will be described.

  The microphone array device 6 according to the sixth embodiment can be realized by, for example, a computer 70 shown in FIG. The computer 70 includes a CPU 72, microphones 11 a and 11 b, a memory 44, a nonvolatile storage unit 46, and a speaker 54, which are connected to each other via a bus 56. The storage unit 46 can be realized by an HDD (Hard Disk Drive), a flash memory, or the like.

  The storage unit 46 as a recording medium stores a noise suppression program 58 for causing the computer 70 to function as the microphone array device 6. The CPU 72 reads out the noise suppression program 58 from the storage unit 46 and develops it in the memory 44, and sequentially executes processes included in the noise suppression program 58.

  The noise suppression program 58 includes a non-stationary noise reduction process 60, a synchronous subtraction coefficient calculation process 62, and a directional sound reception process 64. The CPU 72 operates as the unsteady noise reduction unit 12 illustrated in FIG. 1 by executing the unsteady noise reduction process 60. The CPU 72 operates as the synchronous subtraction coefficient calculator 14 shown in FIG. 1 by executing the synchronous subtraction coefficient calculation process 62. The CPU 72 operates as the directional sound receiving unit 16 shown in FIG. 1 by executing the directional sound receiving process 62. As a result, the computer 70 that has executed the noise suppression program 58 functions as the microphone array device 6.

  Moreover, you may make it operate | move as each function part shown in FIG.5 and FIG.8 by performing each process of the noise suppression program 58. FIG. Further, the noise suppression program 58 may further include a parameter update process, and may operate as each functional unit illustrated in FIG. 12 by executing each process.

  As described above, the microphone array device according to the sixth embodiment is a program for causing a computer to function as each functional unit constituting any one of the microphone array devices according to the first to fifth embodiments. Provided.

  In the above description, the noise suppression program 58 is stored (installed) in the storage unit 46 in advance. However, the present invention is not limited to this. For example, the noise suppression program in the disclosed technology can be provided in a form recorded on a recording medium such as a CD-ROM or a DVD-ROM.

  In the first to sixth embodiments, the directivity sound reception process by synchronous subtraction has been described. However, the present invention can also be applied to the case of directivity sound reception by synchronous addition. Also in the case of synchronous addition, a synchronous addition coefficient is calculated so that the correlation between input signals becomes smaller, and one of the input signals is multiplied by a signal obtained by delaying the other input signal by a predetermined number of samples by the synchronous addition coefficient. What is necessary is just to add a signal.

  All documents, patent applications and technical standards mentioned in this specification are to the same extent as if each individual document, patent application and technical standard were specifically and individually stated to be incorporated by reference. Incorporated by reference in the book.

1, 2, 3, 4, 5, 6 Microphone array device 10, 210, 310, 510 Noise suppression device 11 Microphone array 11a, 11b Microphone 12, 212, 312 Unsteady noise reduction unit 12a, 212a, 312a Correlation coefficient calculation Unit 12b analysis unit 12c reduction unit 12d sound source direction detection unit 14, 214, 314, 514 synchronous subtraction coefficient calculation unit 16, 216, 316 directivity sound reception unit 16a, 216aL, 216aR, 316a delay unit 16b, 216b, 316b multiplication unit 16c, 216c, 316c Subtractor 18L, 18R FFT unit 20 IFFT unit 22 Parameter update unit

Claims (11)

A correlation analyzer for analyzing correlation between a plurality of input signals input from each of the plurality of microphones;
A directional sound receiving unit that performs directional sound reception processing so that the degree of synchronous subtraction or synchronous addition decreases as the correlation analyzed by the correlation analysis unit decreases with respect to the plurality of input signals;
Including a noise suppression device.   The noise suppression apparatus according to claim 1, further comprising: a coefficient calculation unit that calculates a coefficient that decreases as the correlation analyzed by the correlation analysis unit decreases as the degree of the synchronous subtraction or synchronous addition. A reduction unit that reduces the levels of the plurality of input signals according to the magnitude of the correlation analyzed by the correlation analysis unit;
The noise suppression device according to claim 1, wherein the directional sound receiving unit performs the directional sound receiving process on the signal processed by the reduction unit.   The noise suppression device according to claim 1, wherein the correlation is an absolute value of a correlation coefficient between the plurality of input signals. A sound source direction detection unit for detecting a sound source direction in the directivity sound receiving process,
The correlation analysis unit analyzes the correlation for each delay amount of sound arrival between the plurality of microphones corresponding to a sound source direction,
The sound source direction detection unit detects the sound source direction based on a delay amount when the correlation analyzed by the correlation analysis unit becomes maximum,
The said directivity sound-receiving part performs the said synchronous subtraction or synchronous addition so that the sound source direction detected by the said sound source direction detection part may have directivity. Noise suppression device. A frequency domain transforming unit that transforms the plurality of input signals into a frequency domain signal;
The noise suppression apparatus according to any one of claims 1 to 5, wherein the directional sound receiving unit performs the directional sound reception processing for each frequency component.   The noise suppression device according to any one of claims 1 to 6, wherein a degree of the synchronous subtraction or synchronous addition is determined to be smaller as a distance between the plurality of microphones is shorter.   The noise suppression effect is enhanced based on a value indicating the noise suppression effect by the directional sound reception processing represented by the ratio of the level of the output signal to the level of the signal input to the directional sound reception unit. The noise suppression apparatus of any one of Claims 1-7 including the update part which updates the grade of the said synchronous subtraction or synchronous addition. A microphone array in which a plurality of microphones are arranged at predetermined intervals;
The noise suppression device according to any one of claims 1 to 8,
A microphone array apparatus including: A correlation analysis step of analyzing a correlation between a plurality of input signals input from each of the plurality of microphones;
A directional sound reception step for performing directional sound reception processing so that the degree of synchronous subtraction or synchronous addition decreases as the correlation analyzed by the correlation analysis step decreases with respect to the plurality of input signals;
A noise suppression method including: On the computer,
A correlation analysis step of analyzing a correlation between a plurality of input signals input from each of the plurality of microphones;
A directional sound reception step for performing directional sound reception processing so that the degree of synchronous subtraction or synchronous addition decreases as the correlation analyzed by the correlation analysis step decreases with respect to the plurality of input signals;
Noise suppression program for executing processing including