JP6763319B2 - Non-purpose sound determination device, program and method - Google Patents

Description

The present invention relates to a non-purpose sound determination device, a program, and a method for determining the presence or absence of a non-purpose sound (for example, background noise) other than the target sound in, for example, voice processing in a telephone or video conference, or voice recognition processing. Applicable.

In recent years, devices that support various voice processing functions such as voice call functions such as smartphones and car navigation systems and voice recognition functions (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 more severe noise environments than before, such as in a crowded city or in a moving vehicle. Therefore, there is an increasing demand for a voice processing device that can maintain call quality and voice recognition performance even in a noisy environment.

In a conventional voice processing device, when extracting and acquiring a target sound, a process of suppressing a non-purpose sound other than the target sound is performed. As a conventional voice processing device that suppresses non-purpose sounds, for example, there is a technique described in Patent Document 1.

In the apparatus described in Patent Document 1, the input audio signal is subjected to delay subtraction processing to form first and second directivity signals having blind spots in the first and second predetermined directions, and these two directivities are directed. Obtain coherence of sexual signals. Then, in the apparatus described in Patent Document 1, the acquired coherence is compared with the determination threshold value, and whether the input voice signal is a target voice section arriving from the target direction or a non-purpose voice section other than that. Is determined, the gain is set according to the determination result, and the gain is multiplied by the input audio signal to attenuate the non-purpose audio.

Japanese Unexamined Patent Publication No. 2013-182044 Japanese Unexamined Patent Publication No. 2014-106337

By the way, the components usually contained in the non-purpose sound include, for example, background noise (for example, crowds in the city, driving noise of a car, etc.) and disturbing sound (for example, a person other than the user of the voice processing device). It can be roughly divided into (speaking voice).

Since the characteristics and behavior of the background noise and the disturbing sound are completely different, it is sufficient for the conventional voice processing device to switch the processing such as changing the suppression parameter of the disturbing sound depending on the presence or absence of the background noise. No effect was obtained. Further, even when the processed sound is used for voice recognition, sufficient recognition performance cannot be obtained unless the recognition processing and noise reduction characteristics are changed depending on the presence or absence of background noise. Therefore, it is important to be able to accurately determine (detect) the presence of background noise.

However, due to the rapid expansion of the usage environment for audio signal processing, it is difficult to accurately determine the existence of background noise in the presence of various unknown sound sources. Therefore, as a premise for performing advanced signal processing as described above, there is an increasing need for a method capable of accurately determining the presence of background noise.

In view of the above problems, a non-purpose sound determination device, a program, and a method capable of accurately determining the presence or absence of a non-purpose sound (for example, background noise) are desired.

The first aspect of the present invention is (1) to acquire 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 to obtain a difference between the acquired frequency domain input signals for each microphone. Based on this, a front suppression signal generation unit that generates a first front suppression signal having a blind spot on the front, (2) a coherence calculation unit that calculates coherence from input signals obtained from the plurality of microphones, and (3) A first feature amount calculation unit that calculates the correlation coefficient of the first frontal suppression signal and the coherence, and the first feature amount indicating the intensity of positive / negative fluctuation of the amplitude of the correlation coefficient. (4) It is characterized by having a background noise presence determination unit that determines the presence or absence of background noise based on the value of the first feature amount calculated by the first feature amount calculation unit.

The second non-purpose sound determination program of the present invention uses the computer to obtain (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 acquired microphones. Based on the difference between the frequency domain input signals for each frequency domain, the coherence is calculated from the front suppression signal generator that generates the first front suppression signal having a blind spot on the front, and (2) the input signals obtained from the plurality of microphones. The coherence calculation unit, (3) the correlation coefficient of the first frontal suppression signal and the coherence, and the first feature quantity representing the intensity of positive and negative fluctuations of the amplitude gradient of the correlation coefficient are calculated. It functions as a background noise presence determination unit that determines the presence or absence of background noise based on the first feature amount calculation unit and (4) the value of the first feature amount calculated by the first feature amount calculation unit. It is characterized by that.

The third invention is a non-purpose sound determination method used in a non-purpose sound determination device, and includes a front suppression signal generation unit, a coherence calculation unit, a first feature amount calculation unit, and a background noise presence determination unit. (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 the acquired frequency domain input signal for each microphone. A first front suppression signal having a blind spot on the front is generated based on the difference between the above, and (2) the coherence calculation unit calculates coherence from the input signals obtained from the plurality of microphones, and (3) the above. The first feature amount calculation unit calculates the correlation coefficient of the first frontal suppression signal and the coherence, and the first feature amount indicating the intensity of positive / negative fluctuation of the slope of the amplitude of the correlation coefficient. (4) The background noise presence determination unit is characterized in that the presence or absence of background noise is determined based on the value of the first feature amount calculated by the first feature amount calculation unit.

According to the present invention, the presence or absence of non-purpose sound (for example, background noise) can be accurately determined.

It is a block diagram which showed the functional structure of the non-purpose sound determination apparatus which concerns on 1st Embodiment. It is explanatory drawing which showed the arrangement example of the microphone which concerns on 1st Embodiment. It is a figure which showed the characteristic of the directional signal applied by the non-purpose sound determination apparatus which concerns on 1st Embodiment. It is a flowchart (No. 1) which showed the process which the background noise existence determination part which concerns on 1st Embodiment determines the presence or absence of background noise. FIG. 2 is a flowchart (No. 2) showing a process of determining the presence or absence of background noise by the background noise presence determination unit according to the first embodiment. It is a block diagram which shows the whole structure of the non-purpose determination apparatus which concerns on 2nd Embodiment. It is a flowchart (No. 1) which showed about the correlation calculation which concerns on 2nd Embodiment, and the process which the interference sound existence determination part determines the presence or absence of an interference sound. It is a flowchart (No. 2) which showed about the correlation calculation which concerns on 2nd Embodiment, and the process which the interference sound existence determination part determines the presence or absence of an interference sound.

(A) First Embodiment Hereinafter, the first embodiment of the non-purpose sound determination device, program and method according to the present invention will be described in detail with reference to the drawings.

(A-1) Configuration of First Embodiment FIG. 1 is a block diagram showing an overall configuration of a non-purpose sound determination device 1 according to the first embodiment.

The non-purpose sound determination 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 representing the input order of the samples, and is represented by 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 non-purpose sound determination device 1 determines whether or not the input signals captured by the microphones m_1 and m_2 include non-purpose sounds (background noise), and supplies the determination result to a voice processing device (not shown). The voice processing device processes the input signal by using the determination result supplied from the non-purpose sound determination device 1. The processing content performed by the voice processing device on the input signal is not limited. The voice processing device uses the determination result supplied from the non-purpose sound determination device 1 for preprocessing of a communication device such as a video conferencing system or a mobile phone terminal or a voice recognition function. The voice processing device uses, for example, the determination result supplied from the non-purpose sound determination device 1 for suppression processing of the non-purpose sound (for example, background noise).

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

As shown in FIG. 2, in this embodiment, the microphones m_1 and m_2 are arranged so that the two microphones m_1 and m_2 are horizontal to the direction in which the target sound arrives (the direction of the sound source of the target sound). It is assumed that there is. Further, in the following, as shown in FIG. 2, the direction in which the target sound arrives is referred to as the forward direction or the front direction when viewed from the position between the two microphones m_1 and m_1. Further, in the following, as shown in FIG. 2, when referred to as a right direction, a left direction, and a rear direction, each direction when the arrival direction of the target sound is viewed from a position between the two microphones m_1 and m_1 is shown. Explain 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-purpose sound including the disturbing sound comes from the left-right direction (horizontal direction).

The non-purpose sound determination device 1 includes an FFT unit 10, a front suppression signal generation unit 20, a coherence calculation unit 30, a correlation and modGI calculation unit 40, and a background noise presence determination unit 50.

The non-purpose sound determination device 1 may be realized by installing a program (a program including the non-purpose sound determination program according to the embodiment) on a computer having a processor, a memory, or the like, but even in this case, the non-purpose sound determination device 1 is non-purpose. The sound determination device 1 can be functionally shown with reference to FIG. The non-purpose sound determination device 1 may be partially or completely realized in terms of hardware.

The FFT unit 10 receives input signal sequences s1 and s2 from microphones m1 and m2, and performs a fast Fourier transform (or discrete Fourier transform) on the input signals s1 and s2. As a result, the input signals s1 and s2 are represented in the frequency domain. The FFT unit 10 includes an analysis frame FRAME1 (K) and an analysis frame FRAME1 (K) composed of a predetermined N samples (N is an arbitrary integer) from the input signals s1 (n) and s2 (n) when performing the fast Fourier transform. It shall constitute FRAME2 (K). An example of constructing 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 value of K, the older the analysis frame, and the larger the value of K, the newer the analysis frame. Further, in the following operation description, unless otherwise specified, the index representing the latest analysis frame to be analyzed is assumed to be K.
FRAME1 (1) = {s1 (1), s1 (2) ..., s1 (i), ... s1 (n)}
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 to obtain a frequency domain signal X1 (f, K) obtained by Fourier transforming the analysis frame FRAME1 (K) composed of the input signal s1 and an input signal. The frequency domain signal X2 (f, K) obtained by Fourier transforming the analysis frame FRAME2 (K) composed of s2 is acquired. Note that f is an index representing the frequency. Further, (f, K) is not a single value, but is composed of m (m is an arbitrary integer) spectral components of a plurality of frequencies f1 to fm as shown in the following equation (2). It shall be.

The FFT unit 10 supplies the frequency domain signals X1 (f, K) and X2 (f, K) to the front suppression signal generation unit 20 and the coherence calculation unit 30.

Note that X1 (f, K) is a complex number and is composed of a real part and an imaginary part. This also applies to X2 (f, K) and “N (f, K)” described later in the front suppression signal generation unit 20.
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 a process of suppressing the signal component in the front direction for each frequency with respect to the signal supplied from the FFT unit 10. In other words, the front suppression signal generation unit 20 functions as a directional filter that suppresses components in the front direction.

For example, as shown in FIG. 3, the front suppression signal generation unit 20 uses a figure eight-shaped bidirectional filter having a blind spot in the front direction, and a component in the front direction from the signal supplied from the FFT unit 10. Form a directional filter that suppresses.

Specifically, the front suppression signal generation unit 20 is as described in 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. The calculation of the following equation (3) corresponds to the process of forming a figure eight bidirectional filter having a blind spot in the front direction as shown in FIG. 3 described above.
N (f, K) = X1 (f, K) -X2 (f, K) ... (3)

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

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

The coherence calculation unit 30 uses the frequency domain signals X1 (f, K) and X2 (f, K) as signals processed by a filter having a strong directional filter in the left direction (first direction) (hereinafter, “directional signal”). Coherence COH (hereinafter referred to as "directional signal B2 (f)") based on "B1 (f)") and a signal processed by a strong directional filter in the right direction (second direction) (hereinafter referred to as "directional signal B2 (f)"). K) is calculated. It should be noted that the directivity directions of the directivity signal B1 (f) and the directivity signal B2 (f) need to be arbitrary directions other than the front direction (however, the directions of B1 (f) and B2 (f) are different. There is).

The specific calculation process (for example, calculation formula) for calculating the coherence COH (K) is not limited, but for example, the same process as in Patent Document 1 (for example, described in Patent Document 1 (3)). Since equations (calculation processing of equation (7)) can be applied, details will be omitted.

Next, the correlation and the processing of the modGI calculation unit 40 will be described.

First, the correlation and modGI calculation unit 40 calculates the correlation coefficient cor (K) of both from the average front suppression signal AVE_N (K) and the coherence COH (K). The reason (principle) for calculating the correlation coefficient cor (K) will be described later, but simply put, by observing the positive and negative of the correlation coefficient cor (K), even if the target sound is superimposed, the disturbing sound This is because can be easily detected.

Here, it is assumed that the target sound arrives from the front direction of the microphones m_1 and m_2, and the non-purpose sound including the disturbing sound arrives from the left-right direction (horizontal direction). For example, when the microphones m_1 and m_2 are applied to the microphone portion of the receiver of a telephone terminal (for example, a mobile phone terminal, etc.), 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 of the telephone terminal other than the speaker comes from the left-right direction (horizontal direction).

Therefore, for example, when "there is no disturbing sound" and "there is a target sound", the average front suppression signal AVE_N (K) of the front suppression signal N (f, K) has the magnitude of the target sound component. It will be a proportional value. As shown in FIG. 2, the directivity characteristics at the time of generating the average front suppression signal AVE_N (K) (front suppression signal N (f, K)) are "no disturbing sound" and "presence of target sound". Even in this case, the signal component arriving from the front direction is also included. However, as shown in FIG. 2, the directivity characteristics at the time of generating the average front suppression signal AVE_N (K) (front suppression signal N (f, K)) include the signal component arriving from the front direction, but laterally. Very small compared to the directional gain. Therefore, the gain of the front suppression signal N (f, K) when "there is no disturbing sound" and "there is a target sound" is smaller than that when the disturbing sound is present.

In addition, the coherence COH (K) can be said to be simply 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 the case where the correlation between the two directional signals B1 (f) and B2 (f) is small, and conversely, the correlation is large when the coherence COH (K) is large. In other words, the case. When the correlation is small, the direction of arrival of the target sound is largely biased to either the right or the left, or the signal has no clear regularity such as noise even if there is no bias. Further, for example, when the microphones m_1 and m_2 are applied to the microphone portion of the receiver of a telephone terminal (for example, a mobile phone terminal, etc.), the speaker's voice (target voice) arrives from the front, and the disturbing voice is other than the front. There is a strong tendency to arrive from. As described above, the coherence COH (K) is a feature quantity having a close relationship with the arrival direction of the input signal. Therefore, the value of coherence COH (K) tends to be large when "there is no disturbing voice" and "there is a target sound", and when "there is disturbing voice", the coherence COH (K) tends to be large. ) Tends to be smaller.

By arranging the behavior of each of the above values by focusing on the presence or absence of disturbing voice, the presence or absence of disturbing voice can be determined under the following conditions. In the following, a condition that "there is no disturbing sound" and "there is a target sound" (hereinafter referred to as "first condition") and a condition that "there is a disturbing sound" (hereinafter, "second condition"). The method of determining the presence or absence of disturbed voice will be described in each case (referred to as "condition").

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

On the other hand, in the case of the second condition (when "interfering voice is present"), the value of coherence COH (K) tends to be small, and the average front suppression signal AVE_N (K) tends to be large.

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

In the absence of disturbing voice, the correlation coefficient cor (K) tends to be a positive value (a value greater than or equal to a predetermined value indicating that the correlation is high). On the other hand, when the disturbing voice is present, 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), for example, the presence or absence of the disturbing voice is determined (detects the section in which the disturbing voice exists) by a simple process of determining whether the correlation coefficient cor (K) is positive or negative. be able to.

Therefore, the correlation and modGI calculation unit 40 of this embodiment first obtains the correlation coefficient cor (K) and detects the section in which the disturbing voice exists.

The specific calculation method when the correlation and modGI calculation unit 40 obtains the correlation coefficient cor (K) is not limited. For example, the correlation and modGI calculation unit 40 includes the following (5). The correlation coefficient cor (K) may be obtained using an equation. In the following equation (5), Cov [AVE_N (K), COH (K)] indicates the covariance of the average front suppression signal AVE_N (K) and the coherence COH (K). Further, in the following equation (5), σN (f, K) indicates the standard deviation of the average front suppression signal AVE_N (K). Further, in the following equation (5), σCOH (K) indicates the standard deviation of coherence COH (K). When the correlation coefficient cor (K) is calculated by the following equation (5), the standard deviation is used for the results of the predetermined number of i frames processed most recently 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, the i-th frame (K-i-th frame, K- (i-1)) processed most recently. The standard deviation (σN (f, K), and σCOH (K)) and covariance (of the second frame, ..., K-1st frame, Kth frame) are used with COH and AVE_N, respectively. Cov [AVE_N (K), COH (K)]) may be obtained. In other words, the correlation and modGI calculation unit 40 uses the most recently obtained i AVE_N and COH as samples in the process of obtaining the correlation coefficient cor (K), and uses the standard deviation and covariance in the following equation (5). You may want to find the variance.

Then, the correlation and modGI calculation unit 40 obtains the modGI of the correlation by substituting the calculated cor (K) by the formula (5) into the formula (13) of Patent Document 2 (hereinafter, the obtained modGI of the correlation). Is referred to as "cor_modGI (K)"). The reason for calculating the mod GI of the correlation, the usage method, and the like will be described later.

Next, the processing of the background noise presence determination unit 50 will be described.

The background noise presence determination unit 50 determines (detects) a section in which the background noise exists by using the correlation and the cor_modGI (K) obtained by the modGI calculation unit 40.

By the way, in the presence of background noise, the behavior of the correlation (correlation coefficient cor (K)) between the average front suppression signal AVE_N (K) and the coherence COH (K) described above changes as follows.

The macro behavior that the correlation coefficient cor (K) is a negative value when there is an interfering sound and the correlation coefficient cor (K) is a positive value when there is no interfering sound is maintained to some extent. However, under the influence of background noise, the irregularity of fluctuations in the amplitude of the front suppression signal (average front suppression signal AVE_N (K)) increases, whereas the dynamic range of coherence COH (K) is small. To some extent, the irregularity of amplitude does not change drastically. Therefore, the synchrony between the increase / decrease of the frontal suppression signal and the increase / decrease of the coherence COH (K) is impaired, and the fluctuation of the increase / decrease of the correlation coefficient cor (K) becomes large. In addition, the frequency of positive and negative fluctuations of the correlation coefficient cor (K) increases.

From the above, as the influence of background noise increases, the fluctuation of the increase / decrease in the correlation coefficient cor (K) and the frequency of positive / negative fluctuations increase. In this way, when background noise is present, the frequency of fluctuations in the increase / decrease in the correlation coefficient cor (K) and the frequency of positive / negative fluctuations increase, and these fluctuations increase as the influence of the background noise increases. This behavior is derived only from background noise. Therefore, the background noise existence determination unit 50 observes the intensity of the positive / negative fluctuation of the slope of the correlation coefficient cor (K) to determine whether the background noise exists without being affected by the target voice or the disturbing voice. Can be determined.

By the way, modGI (defined by the equation (13) of Patent Document 2) represents the frequency with which the positive and negative of the slope of the waveform fluctuates. The modGI has a feature that the smaller the positive / negative fluctuation of the slope of the signal, the smaller the modGI, whereas the larger the positive / negative fluctuation of the slope, the larger the modGI.

Therefore, the background noise existence determination unit 50 refers to the modGI (cor_modGI (K)) of the above-mentioned correlation and the correlation obtained by the modGI calculation unit 40, and if the cor_modGI (K) is large, the background noise exists, and conversely, the cor_modGI If (K) is small, it can be determined that there is no background noise.

For example, when the background noise presence determination unit 50 of this embodiment is larger than a predetermined threshold value, the background noise presence determination unit 50 outputs a value indicating the presence of background noise (for example, “1”), and the cor_modGI (K) is the threshold value. In the case of the following, a value indicating no background noise (for example, "0") may be output. Various various values can be applied to the threshold value, and the optimum value is determined by, for example, various simulations and statistical analysis.

Further, the background noise presence determination unit 50 outputs a signal Y (K) indicating the determination result. The format of the signal Y (K) is not limited, and for example, a value indicating "with background noise" (for example, "1") or a value indicating "without background noise" (for example, "0") is used. It may be output. The method by which the background noise presence determination unit 50 outputs the signal Y (K) and the supply destination are not limited.

(A-2) Operation of the First Embodiment Next, the operation (determination method of the embodiment) of the non-purpose sound determination device 1 of this embodiment having the above configuration will be described.

First, the overall operation of the non-purpose sound determination device 1 will be described with reference to FIG.

It is assumed that input signals s1 (n) and s2 (n) for one frame (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 Fourier transforms the analysis frames FRAME1 (K) and FRAME2 (K) based on the input signals s1 (n) and s2 (n) for one frame, and performs Fourier transform on the signals X1 (f,) indicated in the frequency domain. 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.

The front suppression signal generation unit 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 the average front suppression signal AVE_N (K) based on the front suppression signal N (f, K) and supplies it to the correlation and modGI calculation unit 40.

On the other hand, the coherence calculation unit 30 generates a coherence COH (K) based on the supplied X1 (f, K) and X2 (f, K) and supplies it to the correlation and modGI calculation unit 40.

The correlation and modGI calculation unit 40 calculates the correlation coefficient cor (K) based on the average frontal suppression signal AVE_N (K) and the coherence COH (K), and based on the calculated correlation coefficient cor (K). The cor_modGI (K) is calculated.

Then, the background noise presence determination unit 50 determines the presence or absence of background noise based on the cor_modGI (K), and outputs the determination result as a signal Y (K).

Next, the operation details of the background noise presence determination unit 50 will be described with reference to the flowcharts of FIGS. 4 and 5.

FIG. 4 is a flowchart showing a process in which the background noise presence determination unit 50 determines the presence or absence of background noise. FIG. 5 is a flowchart showing a part of the processing of the flowchart of FIG. The background noise presence determination unit 50 determines the presence or absence of background noise by processing the flowcharts of FIGS. 4 and 5 each time cor_modGI (K) (data for one frame) is supplied, and outputs a signal Y (K). It shall be output.

When the background noise presence determination unit 50 is supplied with the cor_modGI (K) (S101), the background noise presence determination unit 50 determines the presence or absence of the background noise based on the cor_modGI (K) and the threshold value Θ (S102), and the signal Y indicating the determination result. (K) is generated and output (S103).

Next, a specific example of the determination process performed by the background noise presence determination unit 50 in step S102 will be described with reference to the flowchart of FIG.

When the background noise presence determination unit 50 starts the determination process, it confirms the value of cor_modGI (K) (S201), and determines the presence or absence of disturbing sound according to the value of cor_modGI (K).

Specifically, the background noise presence determination unit 50 determines that "there is background noise" when the cor_modGI (K) is larger than the threshold value Θ (S202), and when the cor_modGI (K) is less than the threshold value Θ, " It is determined that there is no background noise (S203).

(A-3) Effect of First Embodiment As described above, according to the first embodiment, in the non-purpose sound determination device 1, the mod GI of the correlation between the front suppression signal and the coherence is small when background noise is present. Therefore, the presence or absence of non-purpose sound (background noise) can be accurately determined based on the characteristic behavior that the background noise becomes large when there is no background noise. Then, the optimum voice processing can be realized at the supply destination of the determination result depending on the presence or absence of background noise. That is, by applying the determination result of the non-purpose sound determination device 1 of this embodiment to the audio processing of the audio processing device (for example, preprocessing of a communication device such as a video conference system or a mobile phone or a voice recognition function), It is expected that the performance of the voice processing device will be improved (for example, the performance of suppressing non-purpose sounds such as background noise will be improved).

(B) Second Embodiment Hereinafter, a second embodiment of the non-purpose sound determination device, program and method according to the present invention will be described in detail with reference to the drawings.

(B-1) Configuration of Second Embodiment FIG. 6 is a block diagram showing the overall configuration of the non-purpose sound determination device 2 according to the second embodiment, and is the same as and corresponding to FIG. 1 described above. Is the same and is shown with a corresponding code. Each part of the non-purpose determination device of the second embodiment may be configured by hardware, or some configurations may be configured by software.

In FIG. 6, the non-purpose sound determination device 2 has the same FFT unit 10, front suppression signal generation unit 20, coherence calculation unit 30, correlation and modGI calculation unit 40, as in the non-purpose sound determination device 1 shown in FIG. It has a background noise presence determination unit 50, and a correlation calculation and interference sound presence determination unit 60 peculiar to this embodiment.

The non-purpose sound determination device 2 of this embodiment has a correlation calculation and an interfering sound presence determination unit 60 added as compared with the non-purpose sound determination device 1 of FIG. In the following, only the added interfering sound presence determination unit 60 will be described.

The interfering sound presence determination unit 60 correlates the average front suppression signal AVE_N (K) from Long_N (K) obtained by performing long-term averaging processing and coherence COH (K) (correlation coefficient cor_l (K)). Is calculated to determine the presence or absence of disturbing voice. The correlation and modGI calculation unit 40 also calculated the correlation coefficient cor (K), but the interfering sound presence determination unit 60 was not the average front suppression signal AVE_N (K), but Long_N subjected to long-term averaging processing. The difference is that (K) is used. Hereinafter, the reason why Long_N (K) subjected to the long-term averaging treatment is used will be described.

As described above, by introducing the correlation coefficient cor (K), it is possible to determine the presence or absence of disturbed voice by, for example, a simple process of determining whether the correlation coefficient cor (K) is positive or negative. However, as the influence of background noise increases, the fluctuation of the increase / decrease in the correlation coefficient cor (K) and the frequency of positive / negative fluctuations increase. Therefore, the correlation coefficient cor (K) may be negative regardless of the presence or absence of disturbing sound, and an erroneous determination may occur. Therefore, the disturbing sound presence determination unit 60 calculates the correlation (cor_l (K)) using Long_N (K) that has been subjected to long-term average processing that suppresses fluctuations in the average front suppression signal AVE_N (K). Address challenges. This is because, as described above, irregular fluctuations in the front suppression signal due to the influence of background noise are also the cause of changes in the behavior of the correlation.

Specifically, the disturbing sound presence determination unit 60 performs a calculation as shown in the following equation (6) based on the average front suppression signal AVE_N (K) supplied from the front suppression signal generation unit 20, and performs a long-term average. The treated Long_N (K) is generated.
Long_N (K) = λ × AVE_N (K) + (1-λ) × AVE_N (K-1) (0.0 <λ <1.0)… (6)

In the equation (6), the time constant λ is a value having a role of controlling how much the instantaneous input value is reflected with respect to the long-term average value. The larger the time constant λ, the stronger the influence of the instantaneous input, and the smaller the time constant λ, the less the influence of the instantaneous input. Therefore, the smaller the time constant λ is, the more the fluctuation of the front suppression signal can be suppressed, and as a result, the fluctuation of the correlation can be suppressed. The value set in the time constant λ is not limited, but in this embodiment, it is a variable value that varies depending on the determination result of the background noise presence determination unit 50. For example, when the signal Y (K) indicating the presence determination result of the background noise sent by the background noise presence determination unit 50 is a value indicating "with background noise", the disturbing sound presence determination unit 60 has a background. Judging that the influence of noise is large and the fluctuation of the front suppression signal is large, the time constant λ is set to a larger value than that without background noise. On the other hand, when the signal Y (K) has a value indicating "no background noise", the disturbing sound presence determination unit 60 determines that the influence of the background noise is small and the fluctuation of the front suppression signal is small, and the time constant λ Is set to a smaller value than with background noise.

The disturbing sound presence determination unit 60 calculates the correlation coefficient cor_l (K) from the average front suppression signal Long_N (K) that has undergone long-term averaging processing and the coherence COH (K) acquired from the coherence calculation unit 30. Since the method for calculating the correlation coefficient cor_l (K) can be calculated by the same method as the method for calculating the correlation coefficient cor (K) described above, the description thereof is omitted here.

The disturbing sound presence determination unit 60 has no disturbing sound when the correlation coefficient cor_l (K) is larger than 0 (when the correlation coefficient cor_l (K) is 0 or positive; when cor_l (K) ≥ 0). When the correlation coefficient cor_l (K) is less than 0 (when the correlation coefficient cor_l (K) is negative; when cor_l (K) <0), it is determined that there is an interfering voice.

Further, the disturbing sound presence determination unit 60 outputs a signal R (K) indicating the determination result. The format of the signal R (K) is not limited, but for example, a value indicating "with disturbing voice" (for example, "1") or a value indicating "without disturbing voice" (for example, "0") is used. It may be output. The method by which the disturbing sound presence determination unit 60 outputs the signal R (K) and the supply destination are not limited.

(B-2) Operation of the Second Embodiment Next, the operation of the non-purpose sound determination device 2 of the second embodiment having the above configuration (determination method of the embodiment) will be described. Since the operations of each unit other than the correlation calculation and the interference sound presence determination unit 60 are the same as those in the first embodiment, the description thereof will be omitted. In the following, the detailed operations of the correlation calculation and the interference sound presence determination unit 60 are shown. 7. This will be described with reference to the flowchart of FIG.

FIG. 7 is a flowchart showing the correlation calculation and the process of determining the presence / absence of the disturbing sound by the disturbing sound presence determination unit 60. FIG. 8 is a flowchart showing a part of the processing of the flowchart of FIG. 7. The correlation calculation and interference sound presence determination unit 60 determines the presence or absence of interference sound by processing the flowcharts of FIGS. 8 and 7 each time Y (K) (data for one frame) is supplied, and signals R ( K) shall be output.

When the signal Y (K) indicating the determination result of the presence of background noise is supplied (S301), the correlation calculation and the presence determination unit 60 of the disturbing sound confirms the value of the signal Y (K) and sets the time constant λ. The value to be used is controlled (S302).

The correlation calculation and interfering sound presence determination unit 60 performs long-term average processing based on the supplied average front suppression signal AVE_N (K) and the time constant λ whose value is set by the processing of step S302 described above. The front suppression signal Long_N (K) is calculated (S303).

The correlation calculation and the presence determination unit 60 of the disturbing sound calculates the correlation coefficient cor_l (K) based on the supplied coherence COH (K) and the average front suppression signal Long_N (K) subjected to the long-term average processing ( S304).

When the correlation coefficient cor_l (K) is larger than 0 (when the correlation coefficient cor_l (K) is 0 or a positive value; when cor_l (K) ≥ 0), the correlation calculation and interference sound presence determination unit 60 is used. Is judged as "no disturbing sound", and when the correlation coefficient cor_l (K) is less than 0 (when the correlation coefficient cor_l (K) is a negative value; when cor_l (K) <0), "jamming" It is determined that "there is sound" (S305).

The correlation calculation and interference sound presence determination unit 60 generates and outputs a signal R (K) indicating a determination result by the process of step S305 described above (S306).

Next, a specific example of the determination process performed by the correlation calculation and the interference sound presence determination unit 60 in step S302 will be described with reference to the flowchart of FIG.

The background noise presence determination unit 50 confirms the value of the signal Y (K) (“1” indicating “with background noise” or “0” indicating “without background noise”) (S401), and signals Y (K). Is a value indicating "with background noise", a large value is set for the time constant λ (S402), and when the signal Y (K) is a value indicating "without background noise", the time constant is set. A small value is set for λ (S403).

(B-3) Effect of Second Embodiment According to the second embodiment, the following effects are obtained in addition to the effect of the first embodiment.

In the non-purpose sound determination device 2 of the second embodiment, the presence / absence of disturbing sound is determined based on the value of the correlation coefficient cor_l (K). As a result, the non-purpose sound determination device 2 of the second embodiment can accurately determine the presence or absence of the disturbing voice, so that the optimum voice processing can be performed at the supply destination of the determination result according to the presence or absence of the disturbing voice. It can be realized. That is, by applying the determination result of the non-purpose sound determination device 2 of this embodiment to the audio processing of the audio processing device (for example, preprocessing of a communication device such as a video conference system or a mobile phone or a voice recognition function), It is expected that the performance of the voice processing device will be improved (for example, the performance of suppressing non-purpose sounds such as disturbing voices will be improved).

Since the average front suppression signal Long_N (K) used for calculating the correlation coefficient cor_l (K) is subjected to long-term average processing after the optimum time constant λ is set according to the presence or absence of background noise. , It is possible to realize the existence determination processing of the disturbing sound that is robust against the fluctuation of the background noise.

(C) Other Embodiments In addition to the above embodiments, modified embodiments as illustrated below can also be mentioned.

(C-1) In the first embodiment, the case where modGI is applied is shown, but in the GI before modification (Equation (4) of Patent Document 2), the number of times the inclination direction of the signal waveform changes and the number thereof. Since it is an index for measuring the size, GI may be applied instead of modGI in the first embodiment.

(C-2) In the second embodiment, the time constant λ used when calculating the average front suppression signal Long_N (K) may be a preset fixed value.

1, 2 ... Non-purpose sound determination device, 10 ... FFT unit, 20 ... Front suppression signal generation unit, 30 ... Coherence calculation unit, 40 ... modGI calculation unit, 50 ... Background noise presence determination unit, 60 ... Interfering sound presence determination unit , AVE_N, Long_N ... average frontal suppression signal, COH ... coherence, cor, cor_l ... correlation coefficient, Θ ... threshold, λ ... time constant.

Claims (7)

A first frequency domain input signal obtained by converting input signals obtained from a plurality of microphones from a time domain to a frequency domain, and having a blind spot on the front surface based on the difference between the acquired frequency domain input signals for each microphone. Front suppression signal generator that generates the front suppression signal of
A coherence calculation unit that calculates coherence from input signals obtained from the plurality of microphones,
A first feature amount calculation unit for calculating the correlation coefficient of the first frontal suppression signal and the coherence, and a first feature amount indicating the intensity of positive / negative fluctuation of the amplitude gradient of the correlation coefficient. ,
A non-purpose sound determination device including a background noise presence determination unit that determines the presence or absence of background noise based on the value of the first feature amount calculated by the first feature amount calculation unit. The first feature amount calculation unit calculates the modGI value of the correlation coefficient as the first feature amount.
The background noise presence determination unit determines that background noise is present if the first feature amount is larger than a predetermined threshold value, and that background noise is not present if the first feature amount is smaller than a predetermined threshold value. The non-purpose sound determination device according to claim 1, which is characterized. A second feature amount calculation unit that calculates a second feature amount representing the relationship between the second frontal suppression signal obtained by subjecting the first frontal suppression signal to a long-term averaging process and the coherence, and a second feature amount calculation unit.
The invention according to claim 1 or 2, further comprising a disturbing sound presence determination unit that determines the presence or absence of disturbing voice based on the value of the second feature amount calculated by the second feature amount calculation unit. Non-purpose sound judgment device. The second feature amount calculation unit calculates the correlation coefficient between the second front suppression signal and the coherence as the second feature amount.
The non-purpose sound determination device according to claim 3, wherein the disturbing voice presence determination unit determines the presence or absence of disturbing voice based on the positive or negative value of the value of the second feature amount. The second feature amount calculation unit controls the time constant to be subjected to the long-term averaging process applied to the second front suppression signal according to the determination result of the presence / absence of background noise by the background noise presence determination unit. The non-purpose sound determination device according to claim 4, wherein the non-purpose sound determination device is characterized. Computer,
A first frequency domain input signal obtained by converting input signals obtained from a plurality of microphones from a time domain to a frequency domain, and having a blind spot on the front surface based on the difference between the acquired frequency domain input signals for each microphone. Front suppression signal generator that generates the front suppression signal of
A coherence calculation unit that calculates coherence from input signals obtained from the plurality of microphones,
A first feature amount calculation unit for calculating the correlation coefficient of the first frontal suppression signal and the coherence, and a first feature amount indicating the intensity of positive / negative fluctuation of the amplitude gradient of the correlation coefficient. ,
A non-purpose sound determination program characterized in that it functions as a background noise presence determination unit that determines the presence or absence of background noise based on the value of the first feature amount calculated by the first feature amount calculation unit. This is a non-purpose sound determination method used in a non-purpose sound determination device.
It has a front suppression signal generation unit, a coherence calculation unit, a first feature amount calculation unit, and a background noise presence determination unit.
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 based on the difference between the acquired frequency domain input signals for each microphone. Generates a first frontal suppression signal with a frontal blind spot,
The coherence calculation unit calculates coherence from the input signals obtained from the plurality of microphones.
The first feature amount calculation unit obtains the correlation coefficient of the first front suppression signal and the coherence, and the first feature amount representing the intensity of positive / negative fluctuation of the slope of the amplitude of the correlation coefficient. Calculate and
The non-purpose sound determination method, wherein the background noise presence determination unit determines the presence or absence of background noise based on the value of the first feature amount calculated by the first feature amount calculation unit.

Images