JP2018032931A - Acoustic signal processing device, program and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sensitivity calibration gain calculation method capable of accurately calculating a sensitivity calibration gain and more easily setting a threshold even if there of an interference sound.SOLUTION: An acoustic signal processing device 10 comprises: a front face suppression signal generation part 12 for generating a front face suppression signal having a dead angle on a front face based on a difference of multiple frequency domain signals obtained by converting multiple input signals s1(n) and s2(n) obtained from multiple microphones m_1 and m_2 from a time domain to a frequency domain; a coherence calculation part 13 for calculating coherence based on the signal obtained from the multiple input signals; a feature amount calculation part 14 for calculating a feature amount representing a relationship between the front face suppression signal and the coherence; a calibration gain calculation part 15 for detecting a target sound phase that is not influenced by an interference sound, based on the feature amount and calculating a calibration gain for each of input signals in the target sound phase; and calibration gain multiplication parts 16 and 17 for calibrating the corresponding input signals by the calibration gains.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an acoustic signal processing device, a program, and a method, and can be applied to acoustic signal processing used in, for example, a communication device or communication software used for a telephone or a videophone, or preprocessing of voice recognition processing. is there.

  In recent years, devices equipped with various voice processing functions such as a voice call function and a voice recognition function have become widespread, such as smartphones and car navigation systems. However, with the widespread use of these devices, voice processing functions are being used in harsher noise environments than before, such as in crowded streets and running cars. For this reason, there is an increasing demand for signal processing technology that can maintain call sound quality and speech recognition performance even in a noisy environment.

JP 2014-68052 A

Kazuyuki Hiraoka, Gen Hori, “Probability Statistics for Programming”, published by Ohm Co., Ltd., October 23, 2009, p. 178-p. 179

  In recent years, an acoustic signal processing technique using a multi-channel microphone has been realized. However, even if the microphones have the same model number, there is a sensitivity difference, and an accurate acoustic feature cannot be calculated unless the sensitivity difference is calibrated. Previously, methods such as measuring microphone sensitivity in advance and setting a correction gain according to the sensitivity difference, or automatically setting a correction gain that matches the average value by comparing the input level for each channel It is dealt with in. However, since the former takes time and the latter fills not only the difference in microphone sensitivity but also the difference in the acquired input signals, there is a problem that the accuracy of the acoustic feature amount calculated in the latter stage is not guaranteed.

  One method of improving this problem is to calculate the calibration gain by comparing the input levels only in the section of the signal component coming from the front of the microphone among the input signals. If this is a signal coming from the front, the distance between each microphone and the sound source is the same, so the acoustic characteristic difference between the signal components reaching the microphone is very small, and the only characteristic difference that occurs between them is the microphone sensitivity. It is assumed that it can be expected. One of the solutions based on this is the technique described in Patent Document 1. This is because the feature quantity called coherence varies depending on whether the target speaker's voice comes from the front or not, and the microphone sensitivity difference calibration gain is calculated in the signal section where the voice comes from the front. Technology. Note that even if there is a difference in sensitivity between microphones, the coherence maintains the behavior that the magnitude varies depending on whether or not the sound comes from the front, so that the sensitivity difference can be calibrated by this method. (Supplement: Refer to Equation 7 in Patent Document 1 for coherence calculation method)

  However, since the method of Patent Document 1 has a high coherence value even when the speech of another speaker (interfering sound) comes from the left and right simultaneously with the target voice coming from the front of the microphone array, the method arrives from the front. Audio components that are not included are also reflected in the calibration gain. Further, since the sensitivity difference between the microphones is random for each microphone array, it is difficult to optimize the threshold value for detecting the signal section coming from the front, and the target speech section may be erroneously determined.

  Therefore, in order to improve the two problems as described above, there is a need for a sensitivity calibration gain calculation method that can accurately calculate the sensitivity calibration gain even if there is an interfering sound and that can set the threshold more easily. .

  In order to solve the above problems, an acoustic signal processing device according to a first aspect of the present invention is (1) a plurality of frequency domains obtained by converting a plurality of input signals obtained from each of a plurality of microphones from a time domain to a frequency domain. (2) a front suppression signal generation unit that generates a frontal suppression signal having a blind spot on the front based on the signal difference; (2) a coherence calculation unit that calculates coherence based on signals obtained from a plurality of input signals; ) A feature value calculation unit that calculates a feature value indicating the relationship between the front suppression signal and the coherence; and (4) detecting a target sound section that is not affected by the disturbing sound based on the feature value and detecting the target sound. It comprises a calibration gain calculation unit that calculates a calibration gain for each input signal in the section, and (5) a calibration unit that calibrates each corresponding input signal with each calibration gain.

  The acoustic signal processing program according to the second aspect of the present invention is based on a difference between a plurality of frequency domain signals obtained by converting (1) a plurality of input signals obtained from each of a plurality of microphones from a time domain to a frequency domain. A front suppression signal generation unit that generates a front suppression signal having a blind spot in front, (2) a coherence calculation unit that calculates coherence based on signals obtained from a plurality of input signals, and (3) a front suppression signal A feature value calculation unit for calculating a feature value representing the relationship with coherence; and (4) detecting a target sound section that is not affected by the interference sound based on the feature value, and for each input signal of the target sound section. A calibration gain calculation unit that calculates a calibration gain, and (5) each calibration gain functions as a calibration unit that calibrates each corresponding input signal.

  The acoustic signal processing method according to the third aspect of the present invention includes: (1) a plurality of frequency domain signals obtained by converting a plurality of input signals obtained from each of a plurality of microphones from a time domain to a frequency domain; (2) a coherence calculation unit calculates coherence based on signals obtained from a plurality of input signals, and (3) a feature amount calculation unit Then, a feature amount representing the relationship between the front suppression signal and coherence is calculated. (4) The calibration gain calculation unit detects a target sound section that is not affected by the interference sound based on the feature amount, and the target sound is calculated. A calibration gain for each input signal in the section is calculated, and (5) a calibration unit calibrates each corresponding input signal with each calibration gain.

  According to the present invention, it is possible to accurately calculate the sensitivity calibration gain even if there is an interference sound, and to set the threshold value more easily.

1 is a block diagram showing an overall configuration of an acoustic signal processing device according to an embodiment. It is explanatory drawing which shows the directivity characteristic formed in the front suppression signal production | generation part which concerns on embodiment. It is a block diagram which shows the structure of the correlation calculation part which concerns on embodiment. It is a block diagram which shows the structure of the calibration gain calculation part which concerns on embodiment. It is a flowchart which shows the processing operation in the calibration gain calculation part which concerns on embodiment.

(A) Main Embodiment Hereinafter, embodiments of an acoustic signal processing device, a program, and a method according to the present invention will be described in detail with reference to the drawings.

(A-1) Configuration of Embodiment FIG. 1 is a block diagram showing an overall configuration of an acoustic signal processing apparatus 10 according to this embodiment.

  In FIG. 1, an acoustic signal processing apparatus 10 includes a plurality of microphones m_1 and m_2 (two cases are illustrated in FIG. 1), an FFT unit 11, a front suppression signal generation unit 12, a coherence calculation unit 13, and a correlation calculation. Unit 14, calibration gain calculator 15, first calibration gain multiplier 16, and second calibration gain multiplier 17.

  The “feature amount calculation unit” described in the claims includes a correlation calculation unit 14. The “calibration gain calculation unit” includes a calibration gain calculation unit 15. Further, the “calibration unit” includes a first calibration gain multiplication unit 16 and a second calibration gain multiplication unit 17.

  In the acoustic signal processing apparatus 10 illustrated in FIG. 1, components other than the microphones m_1 and m_2 can be realized as software (acoustic signal processing program) executed by the CPU, and the functions of the acoustic signal processing program are illustrated in FIG. Can be expressed as

  The microphone m_1 and the microphone m_2 are arranged apart from each other by a predetermined distance (or an arbitrary distance), and each of the microphone m_1 and the microphone m_2 captures surrounding sounds. Each acoustic signal (input signal) captured by each microphone m_1 and microphone m_2 is converted to an analog / digital (A / D) converter (not shown), and each of the input signals s1 (n) and s2 (n) , FFT unit 11, calibration gain calculation unit 15, first calibration gain multiplication unit 16 and second calibration gain multiplication unit 17. Note that n is an index indicating the input order of samples and is expressed by a positive integer. In the text, it is assumed that the smaller the value of n, the older the input sample, and the larger the value, the newer the input sample.

  The FFT unit 11 receives input signals s1 (n) and s2 (n) from the microphones m_1 and m_2, and performs fast Fourier transform (or discrete Fourier transform) on the input signals s1 (n) and s2 (n). is there. Thereby, the input signals s1 (n) and s2 (n) can be converted from the time domain to the frequency domain. Note that, when performing the fast Fourier transform, the FFT unit 11 includes an analysis frame FRAME1 (K) composed of predetermined N (N is an arbitrary integer) samples from the input signals s1 (n) and s2 (n). Assume that FRAME2 (K) is configured.

An example of configuring FRAME1 from the input signal s1 is illustrated in equation (1). In the following equation (1), K is an index indicating the order of frames, and is expressed as a positive integer. In the following, it is assumed that the smaller the K value, the older the analysis frame, and the larger the K value, the newer the analysis frame. In the following description of the operation, it is assumed that the index K represents the latest analysis frame to be analyzed unless otherwise specified.

  The FFT unit 11 performs a fast Fourier transform process for each analysis frame, thereby performing a frequency domain signal X1 (f, K) obtained by performing a Fourier transform on the analysis frame FRAME1 (K) configured from the input signal s1, and an input A frequency domain signal X2 (f, X) obtained by performing Fourier transform on the analysis frame FRAME2 (K) configured from the signal s2 is acquired. The FFT unit 11 supplies the frequency domain signal X 1 (f, K) and the frequency domain signal X 2 (f, X) to the coherence calculation unit 13 while supplying the front suppression signal generation unit 12.

Here, f is an index representing a frequency. Further, the frequency domain signals X1 (f, K) and X2 (f, K) are not single values, but m (f is an arbitrary number) of a plurality of frequencies f1 to fm as shown in the following equation (2). (Integer) component (spectrum component).

  In the above equation (2), X1 (f, K) is a complex number and consists of a real part and an imaginary part. The same applies to X2 (f, K) and the front suppression signal N (f, K) appearing in the front suppression signal generation unit 12 described later.

  The front suppression signal generation unit 12 performs a process of suppressing the signal component coming from the front direction for each frequency component of the signal from the FFT unit. In other words, the front suppression signal generation unit 12 functions as a directivity filter that suppresses a component in the front direction.

  For example, as shown in FIG. 2, the front suppression signal generation unit 12 suppresses the front direction component from the signal from the FFT unit 11 using a bi-directional filter having a blind spot in the front direction of the figure 8 shape. A directional filter is formed.

Specifically, the front suppression signal generation unit 12 performs a calculation such as the expression (3) based on the signals X1 (f, K) and X2 (f, K) from the FFT unit 11 to obtain frequency components. A front suppression signal N (f, N) is generated for each. The calculation of the following formula (3) corresponds to a process of forming an 8-shaped bi-directional filter having a blind spot in the front direction as shown in FIG.
N (f, K) = X1 (f, K) -X2 (f, K) (3)

  As described above, the front suppression signal generation unit 12 acquires each frequency component of frequencies f1 to fm (power for one frame in each frequency band).

Further, the front suppression signal generator 12 calculates an average front suppression signal AVE_N (K) by averaging the front suppression signals N (f, K) over all frequencies f1 to fm according to the equation (4). To do.

  The coherence calculation unit 14 calculates a coherence COH (K) by forming a highly directional signal in a specific direction included in the frequency domain signals X1 (f, K) and X2 (f, K) from the FFT unit 11. .

  Here, the calculation processing of coherence COH (K) in the coherence calculation unit 14 will be described.

The coherence calculation unit 14 processes the signal B1 (f, K) obtained by processing the frequency domain signals X1 (f, K) and X2 (f, K) with a filter having strong directivity in the first direction (for example, left direction). In addition, the coherence calculation unit 14 generates a signal B2 (f that is processed by a filter having strong directivity in the second direction (for example, right direction) from the frequency domain signals X1 (f, K) and X2 (f, K). , K). An existing method can be applied to the formation method of the signals B1 (f) and B2 (f) having high directivity in a specific direction. Here, the following equation (5) is applied to the first direction. An example in which the signal B1 having high directivity is formed and the signal B2 having high directivity in the second direction is formed by applying the following equation (6) will be described.

  In the above formulas (5) and (6), S is the sampling frequency, N is the FFT analysis frame length, τ is the difference in arrival time of sound waves between the microphone m_1 and the microphone m_2, i is the imaginary unit, and f is the frequency. .

Next, the coherence calculation unit 13 performs the operations shown in the following expressions (7) and (8) on the signals B1 (f) and B2 (f) obtained as described above. Obtain coherence COH (K). Here, B2 (f, K) ^* in the equation (7) is a conjugate complex number of B2 (f, K).

  coef (f, K) is a coherence in a component of an index K having an arbitrary index K (an arbitrary frequency f (any one of frequencies f1 to fm) constituting the analysis frames FRAME1 (K) and FRAME2 (K)). .

  When obtaining coef (f, K), if the directionality of the signal B1 (f) is different from that of the signal B (f), the signals B1 (f) and B2 ( The directivity direction according to f) may be any direction other than the front direction. Further, the method for calculating coef (f, K) is not limited to the above calculation method, and for example, the calculation method described in Patent Document 1 can be applied.

  The correlation calculation unit 14 acquires the average front suppression signal AVE_N (K) from the front suppression signal generation unit 12, acquires the coherence COH (K) from the coherence calculation unit 13, and acquires the average front suppression signal AVE_N (K) and the coherence COH. A correlation coefficient cor (K) with (K) is calculated.

  The significance of the correlation calculation unit 14 calculating the correlation coefficient between the front suppression signal (average front suppression signal) having directivity other than the front direction and coherence will be described.

  Here, it is assumed that there is a sound source that emits a target sound in the front direction of the microphone m_1 and the microphone m_2, and the interference sound comes from a direction other than the front direction (for example, the horizontal direction of the microphone m_1 and the microphone m_2). .

  For example, when “no disturbing voice is present” and “the target sound is present”, the front suppression signal has a signal value proportional to the magnitude of the target sound component. However, as shown in FIG. 2, since the gain in the front direction is smaller than the gain in the horizontal direction, the gain is smaller than that in the case where an interfering sound is present.

  The coherence COH (K) is a feature quantity that has a deep relationship with the arrival direction of the input signal, and is rephrased as a correlation between two signal components. This is because Equation (6) is an equation for calculating a correlation for a certain frequency component, and Equation (7) is an equation for calculating an average of correlation values of all frequency components. Therefore, when the coherence COH (K) is small, the correlation between the two signal components is small. On the contrary, when the coherence COH (K) is large, the correlation between the two signal components is large. Can do. The input signal when the coherence COH (K) is small is a signal arriving from a direction other than the front direction because the arrival direction is greatly biased to either the right or the left. On the other hand, the input signal when the coherence COH (K) is large can be said to be a signal arriving from the front direction with little deviation in the arrival direction.

  Then, if “no interference sound exists” and “the target sound exists”, the coherence COH (K) becomes a large value, “the interference sound exists”, and “the target sound exists”. The coherence COH (K) is a small value.

When the above behavior is organized by focusing on the presence or absence of interfering sounds, the following relationship is obtained.
・ If “no disturbing sound” and “target sound exists”, coherence COH (K) is a large value, and the front suppression signal is a value proportional to the magnitude of the target sound component. When the sound is present, the coherence COH (K) is a small value and the front suppression signal is a large value.

By the way, in the case of the above behavior, if the correlation coefficient between the front suppression signal and the coherence COH (K) is introduced, the following can be said.
・ When “no disturbing sound” is present, the correlation coefficient is a positive value. ・ When “disturbing sound is present”, the correlation coefficient is a negative value.
Therefore, it is possible to determine the presence or absence of the interference sound only by observing the positive / negative of the correlation coefficient between the front suppression signal and the coherence. And using this behavior, when the value of the correlation coefficient between the front suppression signal and coherence is `` positive '', it can be determined that it is a section of only the target sound from the front direction, so it is not affected by the interference sound, The calibration gain for the sensitivity difference between the microphones m_1 and m_2 can be calculated. Also, since the target speech section can be detected simply by observing whether the correlation coefficient value is positive or negative, threshold setting is facilitated unlike the prior art.

  Hereinafter, the calculation processing of the correlation coefficient between the front suppression signal and the coherence in the correlation calculation unit 14 will be described in detail with reference to the drawings.

  FIG. 3 is a block diagram illustrating a configuration of the correlation calculation unit 14 according to the embodiment.

  In FIG. 3, the correlation calculation unit 14 includes a front suppression signal / coherence acquisition unit 31, a correlation coefficient calculation unit 32, and a correlation coefficient output unit 33.

  The front suppression signal / coherence acquisition unit 31 acquires the average front suppression signal AVE_N (K) and the coherence COH (K), and the correlation coefficient calculation unit 32 calculates the average front suppression signal AVE_N (K) and the coherence COH (K). ) To calculate the correlation coefficient cor (K). Then, the correlation coefficient output unit 33 outputs the calculated correlation coefficient cor (K) to the calibration gain calculation unit 15.

Here, the calculation method of the correlation coefficient cor (K) is not limited. For example, the calculation method described in Non-Patent Document 1 can be applied. For example, the correlation coefficient cor (K) is obtained for each frame using the following equation (9). In the following equation (9), Cov [AVE_N (K), COH (K)] indicates the covariance between the average front suppression signal AVE_N (K) and the coherence COH (K). In the following equation (9), σAVE_N (K) indicates the standard deviation of the average front suppression signal AVE_N (K), and σCOH (K) indicates the standard deviation of the coherence COH (K). The correlation coefficient cor (K) obtained in this way takes a value of −1.0 to 1.0.

  The calibration gain calculation unit 15 acquires the correlation coefficient cor (K) from the correlation calculation unit 14, observes the positive / negative of the correlation coefficient cor (K), and in the section where the correlation coefficient cor (K) is “positive”. The calibration gain of microphone m_1 and microphone m_2 is calculated using only the input signal.

  FIG. 4 is a block diagram illustrating a configuration of the calibration gain calculation unit 15 according to the embodiment.

  4, the calibration gain calculation unit 15 includes a correlation coefficient and input signal reception unit 41, a calibration gain calculation execution determination unit 42, a calibration gain calculation unit 43, a calibration gain storage unit 44, and a calibration gain output unit 45.

  The correlation coefficient and input signal acquisition unit 41 acquires the correlation coefficient cor (K) and FRAME1 (K) and FRAME2 (K) that are analysis frames of the input signal from the correlation calculation unit 14.

  The calibration gain calculation execution determination unit 42 determines whether the value of the correlation coefficient cor (K) is “positive” or “negative” in order to determine whether or not to execute the calculation of the calibration gain. . That is, when the value of the correlation coefficient cor (K) is “positive”, the calibration gain calculation execution determination unit 42 determines that the input signal does not include the interference sound, and calculates the calibration gain. It is determined that it is a section to be executed. On the other hand, when the value of the correlation coefficient cor (K) is “negative”, the calibration gain calculation execution determination unit 42 determines that the input signal includes an interference sound and does not calculate the calibration gain. It is determined that it is a section.

  The calibration gain calculation unit 43 calculates calibration gains LEVEL_GAIN_1CH and LEVEL_GAIN_2CH with respect to the sensitivity difference between the microphones m_1 and m_2 according to the determination result by the calibration gain calculation execution determination unit 42.

  When the calibration gain calculation execution determination unit 42 determines that the correlation coefficient cor (K) is “positive”, the calibration gain calculation unit 43 calculates the calibration gains LEVEL_GAIN_1CH and LEVEL_GAIN_2CH. On the other hand, when the calibration gain calculation execution determination unit 42 determines that the correlation coefficient cor (K) is “negative”, the calibration gain calculation unit 43 does not calculate the calibration gain and stores it in the calibration gain storage unit 44. Set the value as the calibration gain.

  Here, a method of calculating the calibration gains CALIB_GAIN_1CH and CALIB_GAIN_2CH by the calibration gain calculation unit 43 will be described.

The calibration gain calculation unit 43 uses the following equations (10, 1), (10, 2), (11), (12, 1), and (12, 2) to calibrate the input signal s1. A gain CALIB_GAIN_1CH and a calibration gain CALIB_GAIN_2CH for the input signal s2 are calculated.

  The expression (10, 1) calculates the average LEVEL_1CH of absolute values of all the components of the current frame (Kth frame) of the input signal s1 (n) captured by the microphone m_1. The measured value LEVEL_1CH can be regarded as a value reflecting the sensitivity of the microphone m_1. Equation (10, 2) calculates the average LEVEL_2CH of absolute values of all the components of the current frame (Kth frame) of the input signal s2 (n) captured by the microphone m_2. The value LEVEL_2CH thus obtained can be regarded as a value reflecting the sensitivity of the microphone m_2.

  For example, the sum of absolute values of the constituent elements of each frame in a predetermined number of frames may be used as values LEVEL_1CH and LEVEL_2CH reflecting microphone sensitivity. For example, the average of the absolute values of all the elements (signal components) constituting the latest P (P ≦ K) frames whose correlation coefficient cor (K) is “positive” is a value reflecting the microphone sensitivity. You may make it use as LEVEL_1CH and LEVEL_2CH. In the latter case, the current frame (K-th frame) is stored by storing the sum of absolute values of the constituent elements of the latest P−1 frames whose correlation coefficient cor (K) is “positive”. ) When the information of FRAME1 (K) and FRAME2 (K) is given, the values LEVEL_1CH and LEVEL_2CH reflecting the microphone sensitivity can be easily calculated. As described above, by calculating the average or total value of the absolute values of long-term signal components, it is possible to calculate a value reflecting the microphone sensitivity while suppressing the influence of instantaneous input signal fluctuations.

  Expressions (10, 1) and (10, 2) are examples of calculation formulas for values that reflect microphone sensitivity, and various other calculation formulas can be applied as described above. However, the calculation formula of the value LEVEL_1CH reflecting the microphone sensitivity of the microphone m_1 and the calculation formula of the value LEVEL_2CH reflecting the microphone sensitivity of the microphone m_2 need to be the same calculation formula.

  Expression (11) calculates the average AVE_LEVEL of the sensitivities LEVEL_1CH and LEVEL_2CH of the two microphones m_1 and m_2 as the target sensitivity of the microphones m_1 and m_2 after calibration. Note that the larger or smaller value of the sensitivity levels LEVEL_1CH and LEVEL_2CH of the two microphones m_1 and m_2 may be set as the target sensitivity.

  The expression (12, 1) is calibrated so that the value obtained by multiplying the sensitivity LEVEL_1CH of the microphone m_1 by the calibration gain CALIB_GAIN_1CH becomes the target sensitivity AVE_LEVEL so that it can be understood by considering the expression in which the denominator LEVEL_1CH on the right side is shifted to the left side. This is a formula for determining the gain CALIB_GAIN_1CH. Similarly, the expression (12, 2) is such that the value obtained by multiplying the sensitivity LEVEL_2CH of the microphone m_2 by the calibration gain CALIB_GAIN_2CH becomes the target sensitivity AVE_LEVEL, as can be understood by considering the expression in which the denominator LEVEL_2CH on the right side is shifted to the left side. The calibration gain CALIB_GAIN_2CH is determined by the following equation.

  The calibration gain storage unit 44 stores calibration gains CALIB_GAIN_1CH (= INIT_GAIN_1CH) and CALIB_GAIN_2CH (= INIT_GAIN_2CH) that are applied when the calibration gain calculation unit 43 does not calculate a calibration gain. As such calibration gains INIT_GAIN_1CH and INIT_GAIN_2CH, a value 1.0 that is not calibrated may be applied, or the latest value calculated by the calibration gain calculator 43 may be applied.

  The calibration gain output unit 45 corresponds to the calibration gains CALIB_GAIN_1CH and CALIB_GAIN_2CH obtained by the calculation of the calibration gain calculation unit 43, or the calibration gains INIT_GAIN_1CH and INIT_GAIN_2CH read from the storage unit 24, respectively. It is something to give to.

  The first calibration gain multiplication unit 16 outputs a post-calibration signal y1 (n) obtained by multiplying the input signal s1 (n) from the microphone m_1 by the calibration gain CALIB_GAIN_1CH.

  The second calibration gain multiplication unit 17 outputs a post-calibration signal y2 (n) obtained by multiplying the input signal s2 (n) from the microphone m_2 by the calibration gain CALIB_GAIN_2CH.

(A-2) Operation of Embodiment Next, the operation of the overall processing and the calculation process of the calibration gain in the acoustic signal processing apparatus 10 according to the embodiment will be described in detail with reference to the drawings.

  Input signals s1 (n) and s2 (n) for one frame are input from the microphones m_1 and m_2 to the FFT unit 11 via an AD converter (not shown).

  The FFT unit 11 performs Fourier transform on the analysis frames FRAME1 (K) and FRAME2 (K) based on the input signals s1 (n) and s2 (n) for one frame, and a signal X1 (f, K) indicated in the frequency domain And X2 (f, K). The signals X1 (f, K) and X2 (f, K) generated by the FFT unit 11 are given to the front suppression signal generation unit 12 and the coherence calculation unit 13.

  The front suppression signal generation unit 12 calculates a front suppression signal N (f, K) having directivity in a direction other than the front direction based on the signals X1 (f, K) and X2 (f, K). Then, the front suppression signal generator 12 generates an average front suppression signal AVE_N (f, K) by averaging the front suppression signals N (f, K) over all frequencies, and this average front suppression signal AVE_N (K ) To the correlation calculation unit 14.

  On the other hand, the coherence calculation unit 13 calculates coherence COH based on the signals X1 (f, K) and X2 (f, K), and provides the coherence COH to the correlation calculation unit 14.

  The correlation calculation unit 14 acquires the average front suppression signal AVE_N (f, K) and the coherence COH, calculates a correlation coefficient cor (K) between the average front suppression signal AVE_N (f, K) and the coherence COH, This correlation coefficient cor (K) is given to the calibration gain calculator 15.

  The calibration gain calculation unit 15 acquires the correlation coefficient cor (K), observes the sign of the correlation coefficient cor (K), and determines the signals s1 (n) and s2 (n) according to the determination result. Calculate the calibration gain for. Further, the calibration gain calculation unit 15 outputs the calibration gain CALIB_GAIN_1CH for the signal s1 (n) to the first calibration gain multiplication unit 16, and outputs the calibration gain CALIB_GAIN_2CH for the signal s2 (n) to the second calibration gain multiplication unit 17. .

  FIG. 5 is a flowchart showing the processing operation in the calibration gain calculator 15.

  The correlation coefficient and input signal acquisition unit 41 acquires the correlation coefficient cor (K) from the correlation coefficient unit 14 and calculates FRAME1 (K) and FRAME2 (K) of the input signals s1 (n) and s2 (n). Obtain (S51).

  Then, the calibration gain calculation execution determination unit 42 determines whether the value of the correlation coefficient cor (K) is positive or negative (S52).

  When the correlation coefficient cor (K) is positive, there is no interfering sound coming from a direction other than the front direction, and it is regarded as a target sound section from the front direction, and the calibration gain calculation unit 43 determines the correlation coefficient cor (K ), FRAME1 (K) and FRAME2 (K), the signals according to the expressions (10, 1), (10, 2), (11), (12, 1), (12, 2) Calibration gains CALIB_GAIN_1CH and CALIB_GAIN_2CH for s1 (n) and signal s2 (n) are calculated (S53). At this time, the calibration gain calculation unit 43 stores the calculated calibration gains CALIB_GAIN_1CH and CALIB_GAIN_2CH in the calibration gain storage unit 44, respectively, and updates the calibration gain stored in the calibration gain storage unit 44.

  When the correlation coefficient cor (K) is negative, it is considered that there is an interference sound coming from a direction other than the front direction, and the calibration gain calculation unit 43 uses the values stored in the calibration gain storage unit 44 as the calibration gains CALIB_GAIN_1CH and It is set as CALIB_GAIN_2CH (S54).

  That is, when initial values INIT_GAIN_1CH and INIT_GAIN_2CH of calibration gains are stored in the calibration gain storage unit 44, INIT_GAIN_1CH is set to CALIB_GAIN_1CH, and INIT_GAIN_2CH is set to CALIB_GAIN_2CH. Alternatively, when the latest calibration gain is stored in the calibration gain storage unit 44, the latest calibration gain stored in the calibration gain storage unit 44 is the current calibration gains CALIB_GAIN_1CH and CALIB_GAIN_2CH.

  Then, the calibration gain output unit 45 outputs the calibration gain CALIB_GAIN_1CH to the first calibration gain multiplication unit 16, and outputs the calibration gain CALIB_GAIN_2CH to the second calibration gain multiplication unit 17 (S55). Then, the calibration gain calculation unit 15 updates the index K (S56), proceeds to S51, and performs the calculation process of the calibration gain of the next index.

  Here, since the calibration gain does not fluctuate once the calibration gain has been calculated once, the calibration gain calculation unit 15 is updated from the middle because it is a waste of calculation amount to constantly update the calibration gain. May be stopped. That is, in the environment where the acoustic signal processing apparatus 10 having the microphones m_1 and m_2 is used, after obtaining the calibration gain for the microphones m_1 and m_2 in the initial stage, it is not necessary to update the steady calibration gain. You may make it perform when calculation of a calibration gain is needed suitably.

  Then, the first calibration gain multiplication unit 16 multiplies the signal s1 (n) by the calibration gain CALIB_GAIN_1CH and outputs a post-calibration signal y1 (n), and the second calibration gain multiplication unit 17 adds the signal s2 (n) to the signal s2 (n). Multiply by the calibration gain CALIB_GAIN_2CH to output a post-calibration signal y2 (n).

(A-3) Effect of Embodiment As described above, according to this embodiment, when there is a disturbing sound that arrives from a direction other than the front direction, the correlation coefficient between the front suppression signal and COH is negative. By using the characteristic behavior that the correlation coefficient between the front suppression signal and COH is positive when no disturbing sound is present, the threshold setting for the designer is not affected by the disturbing sound. An easy microphone sensitivity calibration method can be realized.

  As a result, by applying a process for calculating the calibration gain for the microphones m_1 and m_2 to the preprocessing of various signal processing methods using the microphone array, it is possible to expect improvement in the subsequent voice processing performance.

(B) Other Embodiments Although various modified embodiments have been mentioned in the above-described embodiments, the present invention can also be applied to the following modified embodiments.

  (B-1) In the above-described embodiment, the case where the correlation calculation unit calculates the correlation coefficient as the feature quantity between the front suppression signal and the coherence is exemplified. However, the covariance is calculated as the feature quantity between the front suppression signal and the coherence. Even if this value is calculated, the same effect as in the above-described embodiment can be obtained.

  (B-2) In the above-described embodiment, the acoustic signal processing device according to the present invention is applicable to various devices as long as the device has a voice processing function (for example, voice recognition processing) provided with a plurality of microphones. For example, it can be widely applied to smartphones, tablet terminals, video conference terminals, car navigation systems, call center terminals, robots, devices using sound signals as sensor signals, and the like.

  Further, for example, the acoustic signal processing device of the present invention may be mounted on a device having a communication function, and the device may transmit a post-calibration signal to a server having a predetermined voice processing function through a network.

  Furthermore, for example, a device having a communication function including a plurality of microphones may transmit an input signal of each microphone to a server equipped with the acoustic signal processing device of the present invention through a network. In this case, the server equipped with the acoustic signal processing device can calculate the calibration gain for each input signal according to the correlation coefficient between the front suppression signal and the coherence, as in the above-described embodiment.

  (B-3) In the above-described embodiment, the case where there are two microphones has been illustrated, but the present invention can also be applied to an apparatus that acquires an input signal from each of three or more microphones.

DESCRIPTION OF SYMBOLS 10 ... Acoustic signal processing apparatus, m_1 and m_2 ... Microphone, 11 * .FTT part, 12 ... Front suppression signal generation part, 13 ... Coherence calculation part,
14 ... correlation calculation unit, 31 ... front suppression signal, coherence acquisition unit, 32 ... correlation coefficient calculation unit, 33 ... correlation coefficient output unit,
DESCRIPTION OF SYMBOLS 15 ... Calibration gain calculation part, 41 ... Correlation coefficient and input signal acquisition part, 42 ... Calibration gain calculation execution determination part, 43 ... Calibration gain calculation part, 44 ... Calibration gain memory | storage part, 45 ... Calibration gain output part,
16: first calibration gain multiplication unit, 17: second calibration gain multiplication unit.

Claims (6)

A front suppression signal generation unit that generates a front suppression signal having a blind spot on the front based on a difference between a plurality of frequency domain signals obtained by converting a plurality of input signals obtained from each of a plurality of microphones from a time domain to a frequency domain; ,
A coherence calculator that calculates coherence based on signals obtained from the plurality of input signals;
A feature amount calculation unit that calculates a feature amount representing the relationship between the front suppression signal and the coherence;
A calibration gain calculation unit that detects a target sound section that is not affected by the interference sound based on the feature amount, and calculates a calibration gain for each of the input signals in the target sound section;
An acoustic signal processing apparatus comprising: a calibration unit that calibrates each corresponding input signal with each calibration gain.   The acoustic signal processing apparatus according to claim 1, wherein the feature amount calculation unit calculates a correlation value between the front suppression signal and the coherence as the feature amount.   The acoustic signal processing apparatus according to claim 2, wherein the calibration gain calculation unit detects the target sound section in accordance with the sign of the correlation value between the front suppression signal and the coherence. The calibration gain calculator is
When the correlation value between the front suppression signal and the coherence is positive, it is determined as the target sound section, and a calibration gain for each input signal is calculated.
The acoustic signal processing according to claim 2 or 3, wherein when the correlation value between the front suppression signal and the coherence is negative, it is determined as a sound section including an interfering sound and the calibration gain is not calculated. apparatus. Computer
A front suppression signal generation unit that generates a front suppression signal having a blind spot on the front based on a difference between a plurality of frequency domain signals obtained by converting a plurality of input signals obtained from each of a plurality of microphones from a time domain to a frequency domain; ,
A coherence calculator that calculates coherence based on signals obtained from the plurality of input signals;
A feature amount calculation unit that calculates a feature amount representing the relationship between the front suppression signal and the coherence;
A calibration gain calculation unit that detects a target sound section that is not affected by the interference sound based on the feature amount, and calculates a calibration gain for each of the input signals in the target sound section;
An acoustic signal processing program that causes each calibration gain to function as a calibration unit that calibrates each corresponding input signal. The front suppression signal generation unit generates a front suppression signal having a blind spot on the front based on the difference between a plurality of frequency domain signals obtained by converting a plurality of input signals obtained from each of a plurality of microphones from a time domain to a frequency domain. And
A coherence calculator calculates coherence based on the signals obtained from the plurality of input signals,
A feature amount calculating unit calculates a feature amount representing a relationship between the front suppression signal and the coherence;
A calibration gain calculation unit detects a target sound section that is not affected by the interference sound based on the feature amount, calculates a calibration gain for each input signal in the target sound section,
The acoustic signal processing method, wherein the calibration unit calibrates each corresponding input signal with each calibration gain.

Images