JP2017067990A - Voice processing device, program, and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice processing device that can execute flooring processing so as to suppress a disturbing voice appropriately without excess or deficiency.SOLUTION: A voice processing device of the present invention is characterized by having: a front suppression signal generation unit for acquiring a converted frequency domain input signal from a plurality of microphones, and generating a front suppression signal having a dead angle at the front on the basis of a difference between frequency domain input signals for each microphone; a coherence calculation unit for calculating coherence and a coherence filter coefficient; a feature value calculation unit for calculating the feature value of the front suppression signal with the coherence; a flooring threshold calculation unit for calculating a flooring threshold; and a filter processing unit for performing flooring processing on the coherence filter coefficient applying the calculated flooring threshold, and then suppressing a disturbing voice and acquiring a post-suppression signal.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an audio processing apparatus, program, and method, and can be applied to a process for suppressing non-target sounds (for example, disturbing sounds) other than target sounds, for example, in audio processing in telephone calls, video conferences, etc., and in audio recognition processing. .

  In recent years, devices that support various voice processing functions such as voice call functions and voice recognition functions such as smartphones and car navigation systems (hereinafter, these devices are collectively referred to as “voice processing devices”) have become widespread. ing. However, with the widespread use of these voice processing devices, the voice processing devices have come to be used in harsher noise environments than before, such as in crowded streets and in running cars. For this reason, there is an increasing demand for speech processing devices that can maintain call quality and speech recognition performance even in noisy environments.

  When a target sound is extracted and acquired in a conventional speech processing apparatus, processing for suppressing non-target sounds other than the target sound is performed.

JP 2015-26956 A

  By the way, the components included in the normal non-target sound include, for example, background noise (for example, crowds in the city, driving noise of automobiles, etc.) and disturbing sound (for example, people other than the user of the sound processing device). (Speaking voice). Conventionally, various effective suppression methods have been proposed on the assumption that the background noise has constant frequency characteristics and power. On the other hand, the disturbing voice is a human voice as well as the target voice (voice of the voice processing function user) in addition to non-stationary signal power and frequency characteristics. Therefore, in the conventional speech processing device, when detecting the disturbing speech, it is difficult to determine the presence / absence based on the difference in behavior from the target speech such as background noise. For this reason, when trying to suppress interfering speech with a conventional speech processing device, excessive suppression processing is performed regardless of the presence or absence of interfering speech, resulting in significant distortion in sound quality, or residual components of interfering speech due to insufficient suppression Therefore, there arises a problem that the voice quality and voice recognition performance do not reach predetermined levels.

  Here, in view of the above problem (excessive suppression processing, etc.), the conventional speech processing apparatus “compares the suppression coefficient calculated based on the input signal with a predetermined threshold value, and the suppression coefficient is smaller than the threshold value. May use a flooring process that uses a threshold value as a suppression coefficient without using the suppression coefficient (see Patent Document 1).

  However, if the flooring threshold (hereinafter referred to as the flooring threshold) is too large, the suppression performance is insufficient, and if it is too small, distortion of the target speech increases (that is, the sound quality is affected).

  Therefore, there is a demand for a speech processing apparatus, program, and method that can execute flooring processing that can suppress disturbing speech without excessive or insufficient without affecting sound quality.

  The speech processing apparatus according to the first aspect of the present invention acquires (1) a frequency domain input signal obtained by converting input signals obtained from a plurality of microphones from a time domain to a frequency domain, and the obtained frequency domain input for each microphone. A front suppression signal generator for generating a front suppression signal having a blind spot on the front based on the signal difference; and (2) coherence calculation for calculating coherence and a coherence filter coefficient from input signals obtained from the plurality of microphones. And (3) a feature value calculation unit that calculates a feature value that represents the relationship between the front suppression signal and the coherence, and (4) a flooring threshold value that calculates a flooring threshold value using the feature value. And (5) applying a flooring threshold calculated using the flooring threshold calculation unit to the coherence filter coefficient, After performing grayed process, by suppressing the interference sound included in the input signal, and having a filtering section for acquiring a suppression after signal.

  The audio processing program according to the second aspect of the present invention acquires a frequency domain input signal obtained by converting a computer from (1) an input signal obtained from a plurality of microphones from a time domain to a frequency domain. Based on the difference between the frequency domain input signals, a front suppression signal generator that generates a front suppression signal having a blind spot in front, and (2) calculating coherence and a coherence filter coefficient from input signals obtained from the plurality of microphones. A coherence calculation unit that performs (3) a feature amount calculation unit that calculates a feature amount that represents the relationship between the front suppression signal and the coherence, and (4) calculates a flooring threshold using the feature amount. A flooring threshold value calculation unit; and (5) a flooring threshold value calculated using the flooring threshold value calculation unit for the coherence filter coefficient. By applying, after performing flooring processing, by suppressing the interference sound included in the input signal, characterized in that to function as a filter processing unit that acquires the suppression after signal.

  According to a third aspect of the present invention, there is provided a speech processing method for suppressing interfering speech from input signals obtained from a plurality of microphones, a front suppression signal generator, a coherence calculator, a feature amount calculator, a flooring threshold calculator, and a filter. (1) The front suppression signal generation unit acquires a frequency domain input signal obtained by converting input signals obtained from a plurality of microphones from a time domain to a frequency domain, and acquires the frequency for each acquired microphone. Based on the difference between the region input signals, a front suppression signal having a blind spot in front is generated. (2) The coherence calculation unit calculates coherence and a coherence filter coefficient from the input signals obtained from the plurality of microphones. (3) The feature amount calculation unit calculates a feature amount representing a relationship between the front suppression signal and the coherence, and (4) a flooring threshold value. The output unit calculates the flooring threshold using the feature amount, and (5) the filter processing unit calculates the flooring threshold calculated using the flooring threshold calculation unit for the coherence filter coefficient. The method is characterized in that after applying flooring processing, the interfering voice included in the input signal is suppressed to obtain a post-suppression signal.

  According to the present invention, it is possible to provide an audio processing device, a program, and a method capable of executing a flooring process that can suppress disturbing sound without being excessive or insufficient without affecting sound quality.

It is the block diagram shown about the functional structure of the speech processing unit which concerns on embodiment. It is explanatory drawing shown about the example of arrangement | positioning of the microphone which concerns on embodiment. It is the figure (the 1) shown about the characteristic of the directional signal applied with the audio processing apparatus which concerns on embodiment. It is the figure (the 2) shown about the characteristic of the directional signal applied with the audio processing apparatus which concerns on embodiment. It is the flowchart (the 1) shown about the example of operation | movement of the speech processing unit which concerns on embodiment. It is the flowchart (the 2) shown about the example of operation | movement of the audio processing apparatus which concerns on embodiment.

(A) Main Embodiment Hereinafter, an embodiment of a sound processing apparatus, 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 the overall configuration of the speech processing apparatus 1 of this embodiment.

  The audio processing device 1 acquires input signals s1 (n) and s2 (n) from each of the pair of microphones m_1 and m_2 via an AD converter (not shown). Note that n is an index indicating the input order of samples, and is expressed as a positive integer. In the text, it is assumed that the smaller n is the older input sample, and the larger n is the newer input sample.

  The sound processing device 1 performs processing for suppressing non-target sounds (for example, disturbing sounds) included in input signals captured by the microphones m_1 and m_2. The output format of the audio signal output by the audio processing device 1 is not limited, and may be output as digital audio data of an arbitrary format or may be output as an analog audio signal. In this embodiment, the audio processing device 1 will be described assuming that digital audio data in the PCM (Pulse-code modulation) format or the like is output in units of frames. The audio processing device 1 is used for preprocessing (for example, interference noise suppression processing) used for a communication device such as a video conference system or a mobile phone terminal or a voice recognition function.

  FIG. 2 is an explanatory diagram showing an example of the arrangement of the microphones m_1 and m_2.

  As shown in FIG. 2, in this embodiment, the microphones m_1 and m_2 are such that the plane including the two microphones m_1 and m_2 is perpendicular to the direction in which the target sound arrives (the direction of the target sound source). It is assumed that it is arranged. In the following, as shown in FIG. 2, the arrival direction of the target sound is referred to as the front direction or the front direction when viewed from the position between the two microphones m_1 and m_2. In the following, as shown in FIG. 2, when referring to the right direction, the left direction, and the rear direction, each direction when the arrival direction of the target sound is viewed from the position between the two microphones m_1 and m_2 is shown. It will be explained as a thing. In this embodiment, it is assumed that the target sound comes from the front direction of the microphones m_1 and m_2, and the non-target sound including the disturbing sound comes from the left-right direction (lateral direction).

  The speech processing device 1 includes an FFT unit 10, a front suppression signal generation unit 20, a coherence calculation unit 30, a correlation calculation unit 40, a coherence filter processing unit 50, and an IFFT unit 60.

  The voice processing apparatus 1 may be realized by installing a program (a program including the voice processing program according to the embodiment) in a computer having a processor, a memory, and the like. Specifically, it can be shown using FIG. Note that part or all of the audio processing device 1 may be realized by hardware.

The FFT unit 10 receives input signal sequences s1 and s2 from the microphone m1 and the microphone m2, and performs fast Fourier transform (or discrete Fourier transform) on the input signals s1 and s2. As a result, the input signals s1 and s2 are expressed in the frequency domain. Note that, in performing the fast Fourier transform, the FFT unit 10 includes an analysis frame FRAME1 (K) including predetermined N samples (N is an arbitrary integer) 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 shown in the following equation (1). In the following equation (1), K is an index representing the order of frames and is represented by 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 representing the latest analysis frame to be analyzed is K unless otherwise specified.
FRAME1 (1) = {s1 (1), s1 (2)... S1 (i),.
FRAME1 (K) = {s1 (N × K + 1), s1 (N × K + 2).., S1 (N × K + i),... S1 (N × K + N)} (1)

  The FFT unit 10 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 signal. A frequency domain signal X2 (f, K) obtained by Fourier transforming the analysis frame FRAME2 (K) configured from s2 is acquired. Note that f is an index representing a frequency. Further, (f, K) is not a single value, but is composed of m components (spectrum components) of a plurality of frequencies f1 to fm (m is an arbitrary integer) as shown in the following equation (2). It shall be a thing.

  The FFT unit 10 supplies the frequency domain signals X 1 (f, K) and X 2 (f, K) to the front suppression signal generation unit 20 and the coherence calculation unit 30.

X1 (f, K) is a complex number and is composed of a real part and an imaginary part. The same applies to X2 (f, K) and “N (f, K)” described in the front suppression signal generation unit 20 described later.
X1 (f, K) = {X1 (f1, K), X1 (f2, K),... X1 (fi, K) .., X1 (fm, K)} (2)

  Next, the front suppression signal generation unit 20 will be described.

  The front suppression signal generation unit 20 performs processing for suppressing the signal component in the front direction for each frequency component of the signal supplied from the FFT unit 10. In other words, the front suppression signal generation unit 20 functions as a directivity filter that suppresses a component in the front direction.

  For example, as shown in FIG. 3, the front suppression signal generation unit 20 uses an 8-shaped bi-directional filter having a blind spot in the front direction to generate a component in the front direction from the signal supplied from the FFT unit 10. A directional filter that suppresses the noise is formed.

Specifically, the front suppression signal generation unit 20 is represented by the following equation (3) based on the signals “X1 (f, K)” and “X2 (f, K)” supplied from the FFT unit 10. A calculation is performed to generate a front suppression signal N (f, K) for each frequency component. The calculation of the following equation (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)

Then, the front suppression signal generation unit 20 calculates an average front suppression signal AVE_N (K) by averaging N (f, K) over all frequencies using the following equation (4).

  Next, the process of the coherence calculation unit 30 will be described.

  The coherence calculator 30 has a strong directivity (for example, as shown in FIG. 4A) in the left direction (first direction) for the frequency domain signals X1 (f, K) and X2 (f, K). A signal processed by a filter of directivity (hereinafter referred to as “directivity signal B1 (f)”) and directivity strong in the right direction (second direction) (for example, as shown in FIG. 4B) A coherence COH (K) and a coherence filter coefficient coef (f, K) based on a signal (hereinafter, referred to as “directional signal B2 (f)”) processed by a filter having a unidirectionality). .

  coef (f, K) is a component of an arbitrary frequency f (any one of frequencies f1 to fm) constituting a frame (analysis frame FRAME1 (K) and FRAME2 (K)) having an index value K of an arbitrary index. (Ie, the coherence between the directivity signal B1 (f) and the directivity signal B2 (f)).

  When obtaining COH (K) and coef (f, K), the directionality of the directivity signal B1 (f) and the directivity signal B2 (f) is an arbitrary direction other than the front direction (however, , B1 (f) and B2 (f) need to be in different directions).

  A specific calculation process (for example, a calculation formula) for calculating the coherence COH (K) is not limited, but is similar to, for example, Japanese Patent Application Laid-Open No. 2013-182044 (hereinafter referred to as “Reference Document 1”). Since the above process (for example, the calculation process of Expressions (3) to (7) described in Reference Document 1) can be applied, details are omitted. Also, with respect to the coherence filter coefficient coef (f, K), for example, the calculation processing of Expressions (3) to (6) described in Reference Document 1 can be applied, and thus details thereof are omitted.

  As described above, the coherence calculator 30 supplies the calculated coherence COH (K) to the correlation calculator 40 and the coherence filter coefficient coef (f, K) to the coherence filter processor 50.

  The correlation calculation unit 40 determines the presence / absence of a non-target sound by using the front suppression signal N (f, K) having a directivity other than the front (average front suppression signal AVE_N (K)) and coherence COH (K). A possible correlation coefficient cor (K) is calculated.

  Here, description will be made assuming that the target sound comes from the front direction of the microphones m_1 and m_2, and the non-target sound including the disturbing sound comes from the left-right direction (lateral direction). For example, when the microphones m_1 and m_2 are applied to the microphone portion of the handset of a telephone terminal (for example, a mobile phone terminal), the voice of the speaker (user) as the target sound comes from the front direction of the microphones m_1 and m_2. However, the voice other than the speaker of the telephone terminal comes from the left-right direction (lateral direction).

  Therefore, for example, when “no disturbing voice exists” and “the target sound exists”, the average front suppression signal AVE_N (K) of the front suppression signal N (f, K) has the magnitude of the target sound component. Proportional value. As shown in FIG. 2, the directivity characteristics when generating the average front suppression signal AVE_N (K) (front suppression signal N (f, K)) are “no disturbing speech” and “the target sound exists”. This is because the signal component coming from the front direction is also included. However, as shown in FIG. 2, the directivity characteristic when generating the average front suppression signal AVE_N (K) (front suppression signal N (f, K)) includes a signal component coming from the front direction. Very small compared to the direction gain. In addition, the gain of the front suppression signal N (f, K) when “no disturbing sound exists” and “the target sound exists” is smaller than when the disturbing sound exists.

  The coherence COH (K) can be simply described as a correlation (feature amount) between a signal arriving from the first direction (right direction) and a signal arriving from the second direction (left direction). Therefore, the case where the coherence COH (K) is small is a case where the correlation between the two directivity signals B1 (f) and B2 (f) is small, and conversely, the case where the coherence COH (K) is large is large. In other words. The case where the correlation is small is when the arrival direction of the target sound is greatly deviated to the right or left, or a signal having a clear and regularity such as noise even if there is no deviation. For example, when the microphones m_1 and m_2 are applied to the microphone part of a telephone terminal (for example, a cellular phone terminal), the speaker's voice (target voice) comes from the front and the disturbing voice is other than the front. The tendency to come from is strong. As described above, the coherence COH (K) is a feature amount having a deep relationship with the arrival direction of the input signal. Therefore, the value of coherence COH (K) tends to increase when “no disturbing speech exists” and “the target sound exists”, and when “jamming speech exists”, coherence COH (K ) Tends to be smaller.

  If the behavior of each of the above values is organized by paying attention to the presence or absence of interfering speech, the presence or absence of interfering speech can be determined under the following conditions. In the following, the condition that “no disturbing sound exists” and “the target sound exists” (hereinafter referred to as “first condition”) and the condition that “the disturbing sound exists” (hereinafter referred to as “second sound”). The method for determining the presence / absence of interfering speech will be described for each case.

  In the case of the first condition (“no disturbing sound” and “target sound”), the coherence COH (K) is a relatively large value, and the average front suppression signal AVE_N (K) is The value is proportional to the size of the target sound component.

  On the other hand, in the case of the second condition (when “disturbing speech is present”), the value of coherence COH (K) tends to be a small value, and average front suppression signal AVE_N (K) tends to be a large value.

  Therefore, when the correlation coefficient cor (K) of the average front suppression signal AVE_N (K) and coherence COH (K) is introduced, the relationship between the correlation coefficient cor (K) and the presence or absence of disturbing speech is as follows: Become.

  When no disturbing speech exists, the correlation coefficient cor (K) tends to be a positive value (a value equal to or higher than a predetermined value indicating high correlation). On the other hand, when disturbing speech exists, the correlation coefficient cor (K) tends to be a negative value (a value less than a predetermined value indicating that the correlation is low).

  That is, by introducing the correlation coefficient cor (K) between the average front suppression signal AVE_N (K) and the coherence COH (K), for example, simple processing of determining whether the correlation coefficient cor (K) is positive or negative can be performed. The presence or absence of sound can be determined.

  Therefore, the correlation calculation unit 40 of this embodiment first obtains a correlation coefficient cor (K) for determining the presence or absence of disturbing speech.

  The calculation method used when the correlation calculation unit 40 calculates the correlation coefficient cor (K) is not limited. For example, Reference 2 (Kazuyuki Hiraoka, Gen Hori, “Probability Statistics for Programming”, The calculation method described in Ohm, 2009/10/20) can be applied. The correlation calculation unit 40 may obtain the correlation coefficient cor (K) using, for example, the following equation (5).

In the following equation (5), 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 (5), σAVE_N (K) indicates the standard deviation of the average front suppression signal AVE_N (K). Furthermore, in the following equation (5), σCOH (K) represents a standard deviation of coherence COH (K). When the correlation coefficient cor (K) is obtained by the following equation (5), the standard deviation is obtained using the result of a predetermined number i frames most recently processed for AVE_N (K) and COH (K). Or covariance may be obtained. Specifically, in the process of obtaining the correlation coefficient cor (K) by the following equation (5), for example, i frames (Ki-th frame, K- (i-1) processed most recently) The standard deviations (σN (f, K) and σCOH (K)) and covariances (CON and AVE_N) of the first frame,..., K−1th frame, Kth frame) and covariance ( Cov [AVE_N (K), COH (K)]) may be obtained. In other words, in the process of obtaining the correlation coefficient cor (K), the correlation calculation unit 40 uses the i AVE_N and COH obtained most recently as samples, and calculates the standard deviation and covariance in the following equation (5). You may make it ask.

  Next, processing of the coherence filter processing unit 50 will be described.

  As shown in FIG. 1, the coherence filter processing unit 50 of this embodiment will be described as generating an audio signal in which a disturbing audio component is suppressed with respect to X1 (f, K). Therefore, in this embodiment, the disturbing speech suppression signal O (f, K) is a signal obtained by performing disturbing speech suppression processing (filter processing) on X1 (f, K).

  The coherence filter processing unit 50 may perform a process of suppressing the disturbing sound component for both X1 (f, K) and X2 (f, K). Further, the coherence filter processing unit 50 acquires a signal (for example, an average value of a plurality of signals) obtained by synthesizing X1 (f, K) and X2 (f, K), and a disturbing sound component for the acquired signal. You may make it perform the process which suppresses. Specific processing contents in which the coherence filter processing unit 50 suppresses noise will be described later.

  As described above, the coherence filter processing unit 50 obtains the interfering speech suppression signal O (f, K) for each frequency (frequency f1 to fm) of each frame and supplies it to the IFFT unit 60.

  Next, processing of the IFFT unit 60 will be described.

  The IFFT unit 60 performs a process of converting the supplied O (f, K) from a frequency domain signal to a time domain signal to generate an interfering speech suppression signal o (n). The IFFT unit 60 performs an inverse conversion process on the conversion process performed by the FFT unit 10. In this embodiment, since the FFT unit 10 performs FFT (Fast Fourier Transform), the IFFT unit 60 performs IFFT (Inverse Fourier Transform).

  Next, details of the interfering speech suppression processing performed by the coherence filter processing unit 50 will be described.

  The coherence filter processing unit 50 of this embodiment performs flooring processing on the coherence filter coefficient coef (f, K) based on the above-described correlation coefficient cor (f, K). Then, the coherence filter processing unit 50 multiplies the input signal X1 (f, K) by the coherence filter coefficient coef (f, K) subjected to the flooring process, thereby causing the interference sound suppression signal O (f, K). Get.

By the way, the flooring threshold Θ (K) used in the flooring process is a larger value as the influence of the disturbing voice is smaller, and a smaller value as the influence of the disturbing voice is larger. Is desirable. As described above, the coherence filter processing unit 50 adds the correlation coefficient cor (f, K) that varies depending on the presence / absence of interfering sound to a preset flooring threshold value Θ (K). Control can be realized. The correlation calculation unit 40 may obtain the flooring threshold Θ (K) using, for example, the following equation (6). In the following equation (6), Ψ (K) is a predetermined constant.
Θ (K) = Ψ (K) + cor (K) (6)

  The coherence filter processing unit 50 performs flooring processing on the coherence filter coefficient coef (f, K) using the generated flooring threshold Θ (K) (details will be described in the operation section).

Then, the coherence filter processing unit 50 uses the coherence filter coefficient coef (f, K) that has been subjected to the flooring process, and uses the interference signal (interference sound component) of the input signal (X1 (f, K) in this embodiment). ) To suppress the disturbing voice suppression signal O (f, K). In the example of this embodiment, the coherence filter processing unit 50 multiplies the input signal X1 (f, K) by the coherence filter coefficient coef (f, K) for each frequency component as shown in the following equation (7). Thus, the interfering voice suppression signal O (f, K) can be obtained.
Interfering voice suppression signal O (f, K) = input signal X1 (f, K) × coherence filter coefficient coef (f, K)) (7)

(A-2) Operation | movement of embodiment Next, operation | movement of the audio | voice processing apparatus 1 of this embodiment which has the above structures is demonstrated.

  First, the overall operation of the speech processing apparatus 1 will be described with reference to FIG.

  Assume that input signals s1 (n) and s2 (n) for one frame (for one processing unit) are supplied to the FFT unit 10 from each of the microphones m_1 and m_2 via an AD converter (not shown). Then, the FFT unit 10 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 the signal X1 (f, K) and X2 (f, K) are acquired. Then, the signals X1 (f, K) and X2 (f, K) generated by the FFT unit 10 are supplied to the front suppression signal generation unit 20 and the coherence calculation unit 30. Further, the signal X 1 (f, K) generated by the FFT unit 10 is supplied to the coherence filter processing unit 50.

  The front suppression signal generator 20 calculates the front suppression signal N (f, K) based on the supplied X1 (f, K) and X2 (f, K). Then, the front suppression signal generation unit 20 calculates an average front suppression signal AVE_N (K) based on the front suppression signal N (f, K) and supplies the average front suppression signal AVE_N (K) to the correlation calculation unit 40.

  On the other hand, the coherence calculation unit 30 generates a coherence COH (K) and a coherence filter coefficient coef (f, K) based on the supplied X1 (f, K) and X2 (f, K). Then, the coherence calculation unit 30 supplies the generated coherence COH (K) to the correlation calculation unit 40 and the coherence filter coefficient coef (f, K) to the coherence filter processing unit 50.

  The correlation calculation unit 40 calculates a correlation coefficient cor (K) based on the average front suppression signal AVE_N (K) and the coherence COH (K), and supplies the correlation coefficient cor (K) to the coherence filter processing unit 50.

  The coherence filter processing unit 50 calculates the flooring threshold Θ (K) from the supplied correlation coefficient cor (K). The coherence filter processing unit 50 performs a flooring process on the supplied coherence filter coefficient coef (f, K) using the calculated flooring threshold Θ (K). Then, the coherence filter processing unit 50 uses the coherence filter coefficient coef (f, K) subjected to the flooring process to suppress the disturbing sound (disturbing sound component) of the input signal X1 (f, K), The interfering voice suppression signal O (f, K) is generated and supplied to the IFFT unit 60.

  The IFFT unit 60 performs inverse Fourier transform (IFFT) processing on the supplied interfering speech suppression signal O (f, K) to convert it to a disturbing speech suppression signal o (n) in the time domain and outputs it.

  Next, details of the operation of the coherence filter processing unit 50 will be described with reference to the flowcharts of FIGS.

  FIG. 5 is a flowchart showing the processing of the coherence filter processing unit 50. FIG. 6 is a flowchart showing a part of processing (flooring processing) in the flowchart of FIG. Each time the coherence filter processing unit 50 is supplied with the correlation coefficient cor (K), the coherence filter coefficient coef (f, K), and the input signal X1 (f, K), the processing of the flowcharts of FIGS. And the disturbing voice suppression signal O (f, K) is output.

  First, it is assumed that the correlation coefficient cor (K), the coherence filter coefficient coef (f, K), and the input signal X1 (f, K) are supplied to the coherence filter processing unit 50 (S101).

  Next, the coherence filter processing unit 50 calculates a flooring threshold Θ (K) based on the correlation coefficient cor (K) (S102). Specifically, the coherence filter processing unit 50 can obtain the flooring threshold Θ (K) using the above equation (6).

  Next, the coherence filter processing unit 50 performs a flooring process on the coherence filter coefficient coef (f, K) using the calculated flooring threshold Θ (K) (S103).

  Next, the coherence filter processing unit 50 uses the coherence filter coefficient coef (f, K) subjected to the flooring process to suppress the interference sound component of the input signal X1 (f, K) (filter processing). To generate a disturbing voice suppression signal O (f, K) (S104). Specifically, the coherence filter processing unit 50 performs the coherence filter coefficient coef () obtained by performing flooring processing on the input signal X1 (f, K) for each frequency component, as in the above equation (7). By multiplying (multiplying) f, K), the interfering voice suppression signal O (f, K) is obtained.

  Next, the coherence filter processing unit 50 outputs (transmits to the IFFT unit 60) the obtained interfering voice suppression signal O (f, K) (S105).

  Next, a specific example of the flooring process performed by the coherence filter processing unit 50 in step S103 described above will be described with reference to the flowchart of FIG.

  When the flooring process is started, the coherence filter processing unit 50 checks the values of the coherence filter coefficient coef (f, K) and the flooring threshold value Θ (K) (S201), and compares the magnitudes of both values.

  When the coherence filter coefficient coef (f, K) is smaller than the flooring threshold Θ (K), the coherence filter processing unit 50 sets the value of the coherence filter coefficient coef (f, K) as the flooring threshold Θ (K). (S202).

  On the other hand, when the coherence filter coefficient coef (f, K) is equal to or larger than the flooring threshold Θ (K), the coherence filter processing unit 50 keeps the value of the coherence filter coefficient coef (f, K) as it is (especially nothing is processed). (S203).

(A-3) Effects of Embodiment According to this embodiment, the following effects can be achieved.

  In the speech processing apparatus 1 of this embodiment, when the disturbing speech exists, the correlation coefficient cor (K) between the average front suppression signal AVE_N (K) and the coherence COH (K) is negative, and the disturbing speech exists. If not, the flooring process is performed on the coherence filter coefficient coef (f, K) based on the positive characteristic behavior, so that the accuracy of interfering speech suppression can be improved. Thereby, in the voice processing apparatus 1 of this embodiment, the disturbing voice can be suppressed without excess or deficiency without affecting the sound quality with respect to the input signal. That is, in the audio processing of the audio processing device 1 (for example, pre-processing of a communication device such as a video conference system or a cellular phone or a voice recognition function), the performance is improved (for example, the suppression performance of non-target sounds such as disturbing sounds is improved) Can be expected.

(B) Other Embodiments The present invention is not limited to the above-described embodiments, and may include modified embodiments as exemplified below.

  (B-1) The speech processing apparatus 1 according to the above embodiment calculates the flooring threshold Θ (K) by adding the correlation coefficient cor (K) (for example, equation (6)). Not limited to this, an arbitrary calculation may be performed as long as a desired flooring characteristic can be obtained according to the contribution of the disturbing sound.

  (B-2) The speech processing apparatus 1 according to the above embodiment calculates the flooring threshold Θ (K) in units of frames. However, the present invention is not limited to this, and for example, the flooring threshold Θ (K) for each frequency bin. You may ask for. In this case, the speech processing apparatus 1 may calculate the correlation coefficient cor (K) that is the basis of the flooring threshold Θ (K) for each frequency bin.

  (B-3) The audio processing apparatus 1 of the above embodiment has been described with respect to an example in which processing is performed based on input signals supplied from two microphones. However, the audio processing apparatus 1 is supplied from three or more microphones. The determination process may be performed based on the input signal. For example, in the audio processing device 1, based on input signals supplied from three or more microphones, a front suppression signal N (f, K) having a blind spot in the front direction or directivity in a predetermined direction other than the front is provided. The directivity signals B1 (f) and B2 (f) may be acquired and the same processing as in the above embodiment may be performed. That is, in the speech processing apparatus 1, the configuration of the microphone for obtaining the front suppression signal N (f, K) and the directivity signals B1 (f) and B2 (f) is not limited.

  (B-4) In the coherence filter processing unit 50 of the above-described embodiment, the average front suppression signal AVE_N (K) is used as a feature amount representing the relationship between the average front suppression signal AVE_N (K) and the coherence COH (K). Although the correlation coefficient cor (K) with the coherence COH (K) is applied, other types of values may be applied as feature quantities. For example, in the coherence filter processing unit 50, as the feature quantity indicating the relationship between the average front suppression signal AVE_N (K) and the coherence COH (K), the average front suppression signal AVE_N (K) and the coherence COH (K) are combined. You may make it apply dispersion | distribution.

  DESCRIPTION OF SYMBOLS 1 ... Speech processing apparatus, 10 ... FFT part, 20 ... Front suppression signal generation part, 30 ... Coherence calculation part, 40 ... Correlation calculation part, 50 ... Coherence filter processing part, 60 ... IFFT part, m_1, m_2 ... Microphone.

Claims (6)

Obtaining a frequency domain input signal obtained by converting input signals obtained from a plurality of microphones from a time domain to a frequency domain, and based on the obtained difference of the frequency domain input signals for each microphone, frontal suppression having a blind spot in front A front suppression signal generator for generating a signal;
A coherence calculation unit for calculating coherence and a coherence filter coefficient from input signals obtained from the plurality of microphones;
A feature amount calculating unit that calculates a feature amount representing a relationship between the front suppression signal and the coherence;
A flooring threshold value calculation unit that calculates a flooring threshold value using the feature amount;
Applying the flooring threshold calculated by using the flooring threshold calculation unit to the coherence filter coefficient, performing flooring processing, and suppressing the disturbing voice included in the input signal to suppress And a filter processing unit that acquires a post signal.   The audio processing apparatus according to claim 1, wherein the filter processing unit obtains the post-suppression signal by multiplying the input signal by a coherence filter coefficient subjected to flooring processing. The feature amount is a correlation coefficient between the front suppression signal and the coherence,
The voice according to claim 1 or 2, wherein the flooring threshold value calculation unit calculates the flooring threshold value by a predetermined calculation process using a predetermined value and a feature amount. Processing equipment.   The audio processing apparatus according to claim 3, wherein the predetermined value is a positive constant, and the calculation process is an addition process. Computer
Obtaining a frequency domain input signal obtained by converting input signals obtained from a plurality of microphones from a time domain to a frequency domain, and based on the obtained difference of the frequency domain input signals for each microphone, frontal suppression having a blind spot in front A front suppression signal generator for generating a signal;
A coherence calculation unit for calculating coherence and a coherence filter coefficient from input signals obtained from the plurality of microphones;
A feature amount calculating unit that calculates a feature amount representing a relationship between the front suppression signal and the coherence;
A flooring threshold value calculation unit that calculates a flooring threshold value using the feature amount;
Applying the flooring threshold calculated by using the flooring threshold calculation unit to the coherence filter coefficient, performing flooring processing, and suppressing the disturbing voice included in the input signal to suppress An audio processing program that functions as a filter processing unit that acquires a post-signal. In a speech processing method for suppressing interfering speech from input signals obtained from a plurality of microphones,
A front suppression signal generation unit, a coherence calculation unit, a feature amount calculation unit, a flooring threshold calculation unit, and a filter processing unit,
The front suppression signal generation unit obtains a frequency domain input signal obtained by converting input signals obtained from a plurality of microphones from a time domain to a frequency domain, and based on the obtained frequency domain input signal difference for each microphone. Generate a frontal suppression signal with a blind spot in front,
The coherence calculation unit calculates coherence and a coherence filter coefficient from input signals obtained from the plurality of microphones,
The feature amount calculation unit calculates a feature amount representing a relationship between the front suppression signal and the coherence,
The flooring threshold calculation unit calculates a flooring threshold using the feature amount,
The filter processing unit applies a flooring threshold calculated using the flooring threshold calculation unit to the coherence filter coefficient, performs a flooring process, and then includes disturbing speech included in the input signal. And a signal after suppression is acquired.

Images