JP6772890B2 - Signal processing equipment, programs and methods - Google Patents

Description

The present invention relates to a signal processing device, a program, and a method, for example, a communication terminal, an audio device, and a voice that want to emphasize and collect sound sources existing in a specific direction in an environment where a plurality of sound sources exist. It can be applied to recognition devices and the like.

As a technique for extracting a certain target sound source in an environment where multiple sound sources exist, sound source separation using multiple microphones and a beam former using a microphone array in which microphones are arranged on a straight line, a plane, a spherical surface, etc. And nullformers. In particular, when sound sources other than the target sound source are non-stationary, or when there are multiple sound sources, it is difficult to extract the target sound source with a noise suppressor using a single microphone, and it is essential to use two or more microphones. Become.

The beam former using the microphone array described above is a technique for emphasizing and collecting sound only in a specific direction. Beam former is a technique for forming directivity by utilizing the time difference of signals arriving at each microphone.

There are two types of beam formers: addition type and subtraction type. Compared to the additive beamformer, the subtractive beamformer has the advantage that sharp directivity can be formed with a smaller number of microphones.

FIG. 13 is a block diagram showing a configuration related to a subtraction type beamformer when the number of microphones is two. The subtraction type beamformer of FIG. 13 is composed of a first microphone M1, a second microphone M2, a first delay means 3, a second delay means 4, and a subtraction means 5. The first input signal picked up by the first microphone M1 is given to the first delay means 3, and the second input signal picked up by the second microphone M2 is given to the second delay means 4. When the disturbing sound comes from the first microphone M1 side, the first delay means 3 delays the first input signal, so that the disturbing sound included in the first input signal and the second input signal To match the phase of. On the other hand, when the disturbing sound comes from the second microphone M2 side, the second delay means 4 delays the second input signal to match the phase of the disturbing sound. The first delay signal obtained from the first delay means 3 and the second delay signal obtained from the second delay means are given to the subtraction means 5. The subtracting means 5 obtains the emphasized voice by subtracting the second delay signal from the first delay signal. As described above, the subtraction type beamformer emphasizes the target sound by matching the phases of the disturbing sounds contained in the first input signal and the second input signal, subtracting them, and suppressing the disturbing sounds. The subtraction type beam former requires information on the direction of arrival of the disturbing sound given in advance.

By the way, the subtraction type beam former has a problem that the suppression performance of the disturbing sound is greatly deteriorated if the disturbing sound source moves even a little.

FIG. 14 is an explanatory diagram showing an example of emphasizing the voice of the driver U1 in the automobile (vehicle) A by using the conventional signal processing device Z.

For example, in a car navigation system or the like that can be operated by voice using voice recognition as shown in FIG. 14, it is necessary to extract only the driver's voice in the car.

Therefore, when a person is in the driver's seat and the passenger's seat, it is necessary to suppress the voice (interfering sound) of the passenger U2 in the passenger's seat, but the assistant U2 makes a face (interfering sound source) in front, back, left and right. When moved, the subtractive beamformer cannot suppress the disturbing sound.

The Minimum Variance Beamformer (MVB), which is one of the representative adaptive beamformers, is a method that can efficiently suppress disturbing sounds by giving the direction of arrival of the target sound in advance. The MVB suppresses the disturbing sound by minimizing the dispersion of the emphasized sound under the constraint condition that the gain is 1 with respect to the arrival direction of the target sound.

Further, by using the spectrum subtraction method, it is possible to form a strong directivity in the direction of arrival of the target sound source. In Non-Patent Document 1, assuming that the target sound source is always in front, a subtraction type beam former first obtains a target sound suppression signal that suppresses a target sound coming from the front direction, and secondly, a first input. By subtracting the amplitude spectrum of the target sound suppression signal from the amplitude spectrum of the signal (spectrum subtraction), the amplitude spectrum of the emphasized sound that emphasizes the target sound is obtained, and thirdly, the amplitude spectrum of the emphasized sound and the phase of the first input signal are obtained. The enhanced sound is obtained using the spectrum.

Takashi Yato, Makoto Morito, Kei Yamada, Tetsuji Ogawa, "Sound Source Separation Technology by Square Microphone Array", Information Processing, Vol. 51, No. 11, 2010

However, conventional techniques have the following problems.

FIG. 15 is an explanatory diagram showing an image of a target sound and an interfering sound in the automobile A.

The MVB can suppress only one less disturbing sound than the number of microphones. Therefore, when the target sound is emphasized by two microphones as shown in FIG. 14, since the disturbing sound propagates as shown in FIG. 15 (b), the MVB can suppress the direct sound of the disturbing sound but cannot suppress the reflected sound. Therefore, the target sound cannot be sufficiently emphasized.

The technique described in Non-Patent Document 1 suppresses all sounds coming from other than the front direction, even if they are derived from the target sound. Therefore, when the target sound is emphasized by two microphones as shown in FIG. 14, the target sound propagates as shown in FIG. 15 (a). Therefore, the technique described in Non-Patent Document 1 also includes the reflected sound of the target sound. Since it is suppressed, the sound quality of the target sound deteriorates.

Therefore, it is possible to provide a signal processing device, a program, and a method for emphasizing a target sound with less calculation cost and less distortion.

The first signal processing device of the present invention includes (1) a first frequency analysis means for frequency-analyzing a first input signal input from a first sound collecting device to obtain a first input spectrum, and (1). 2) A second frequency analysis means for obtaining a second input spectrum by frequency-analyzing a second input signal input from the second sound collecting device, and (3) obtained by the first frequency analysis means. Based on the first input spectrum and the second input spectrum obtained by the second frequency analysis means, a straight line connecting the position of the first sound collecting device and the position of the second sound collecting device. Calculate the first feature amount in which the values in the front direction and the direction on the first sound collecting device side are made larger and the values in the direction on the second sound collecting device side are made smaller than the vertical front direction. The feature amount calculating means to be used, (4) the filter determining means for obtaining an emphasis filter by mapping the first feature amount calculated by the feature amount calculating means with a predetermined broad-sense monotonous increase function, and (5) the above. It is characterized by including a multiplication means for obtaining an emphasis spectrum by multiplying the first input spectrum obtained by the first frequency analysis means by the enhancement filter obtained by the filter determination means.

The second signal processing program of the present invention is a first frequency analysis means for obtaining a first input spectrum by frequency-analyzing a computer with (1) a first input signal input from a first sound collecting device. And (2) a second frequency analysis means for obtaining a second input spectrum by frequency analysis of the second input signal input from the second sound collecting device, and (3) the first frequency analysis means. Based on the first input spectrum obtained in the above and the second input spectrum obtained by the second frequency analysis means, the position of the first sound collecting device and the position of the second sound collecting device are connected. The first feature that the value in the front direction and the direction on the first sound collecting device side is large and the value in the direction on the second sound collecting device side is small with respect to the front direction perpendicular to the straight line. A feature amount calculating means for calculating an amount, and (4) a filter determining means for obtaining an emphasis filter by mapping the first feature amount calculated by the feature amount calculating means with a predetermined broad-sense monotonous increase function. 5) A multiplication means obtained by multiplying the first input spectrum obtained by the first frequency analysis means by the emphasis filter obtained by the filter determination means to obtain an emphasis spectrum, and (6) the multiplication means obtained by the multiplication means. It is characterized in that it functions as a waveform restoration means for inputting an emphasis spectrum and restoring a signal waveform to obtain an emphasized sound.

The third signal processing method of the present invention includes (1) a first frequency analysis means, a second frequency analysis means, a feature amount calculation means, a filter determination means, and a multiplication means in the signal processing method. 2) The first frequency analysis means obtains a first input spectrum by frequency-analyzing the first input signal input from the first sound collecting device, and (3) the second frequency analysis means. (4) The feature amount calculation means is obtained by the first frequency analysis means by frequency-analyzing the second input signal input from the second sound collecting device to obtain a second input spectrum. A straight line connecting the position of the first sound collecting device and the position of the second sound collecting device based on the obtained first input spectrum and the second input spectrum obtained by the second frequency analysis means. A first feature amount in which the values in the front direction and the direction on the first sound collecting device side are large and the values in the direction on the second sound collecting device side are small with respect to the front direction perpendicular to the frequency. The filter determining means (5) maps the first feature amount calculated by the feature amount calculating means with a predetermined broad monotonous increase function to obtain an emphasis filter, and (6) the multiplication. The means is characterized in that the enhancement spectrum is obtained by multiplying the first input spectrum obtained by the first frequency analysis means by the enhancement filter obtained by the filter determining means.

According to the present invention, it is possible to provide a signal processing device, a program and a method for emphasizing a target sound with less calculation cost and less distortion.

It is a block diagram which showed the functional structure of the signal processing apparatus which concerns on 1st Embodiment. It is explanatory drawing which showed the example of the use environment of the signal processing apparatus which concerns on 1st Embodiment. An example of the feature quantity _Fcenter processed by the signal processing apparatus according to the first embodiment is shown. An example of the feature quantity F _side processed by the signal processing apparatus according to the first embodiment is shown. It is a graph which showed the relationship between the arrival direction θ of the sound processed by the signal processing apparatus which concerns on 1st Embodiment, and DOA feature quantity F. It is a graph which showed the example of the broad sense monotonous increase function processed by the signal processing apparatus which concerns on 1st Embodiment. It is a graph which showed the example of the emphasis filter used in the signal processing apparatus which concerns on 1st Embodiment. It is a graph which showed the example of the emphasis filter G obtained by the filter determination means which concerns on 1st Embodiment. It is a graph which showed the example of the emphasis filter G obtained by the filter determination means which concerns on 2nd Embodiment. It is a graph which showed the comparison between the emphasis filter G obtained by the filter determination means which concerns on 2nd Embodiment, and the enhancement filter G obtained by the filter determination means which concerns on 3rd Embodiment. It is a graph which showed the relationship between the arrival direction θ of the sound processed by the signal processing apparatus which concerns on 4th Embodiment, and DOA feature quantity F'. It is explanatory drawing which showed the example of the emphasis filter G obtained by the filter determination means 404 which concerns on 4th Embodiment. It is a block diagram which shows the structure which concerns on the subtraction type beam former when the number of conventional microphones is two. It is explanatory drawing which showed the example which emphasizes the voice of a driver in an automobile by using the conventional signal processing apparatus. It is explanatory drawing which showed the image of the target sound and the disturbing sound in an automobile.

(A) First Embodiment Hereinafter, the first embodiment of the signal processing apparatus, 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. 2 is an explanatory diagram showing an environment in which the signal processing device 100 according to the first embodiment is used. In FIG. 2, the reference numerals in parentheses are the reference numerals used in the second to fourth embodiments described later.

The signal processing device 100 according to the first embodiment is an explanatory diagram showing an example of emphasizing the voice of the driver U1 in the automobile A. In the automobile A, the driver U1 is seated in the driver's seat, and the passenger U2 is seated in the passenger seat. In the automobile A, the first microphone M1 and the second microphone M2 constituting the microphone array are arranged in front of the driver U1 (front of the driver's seat). The first microphone M1 is arranged on the left side (the side opposite to the assistant U2) and the second microphone M2 is arranged on the right side (the side opposite to the assistant U2) when viewed from the driver U1.

FIG. 1 is a block diagram showing a functional configuration of the signal processing device 100 according to the first embodiment.

The signal processing device 100 of the first embodiment includes a first frequency analysis means 101, a second frequency analysis means 102, a feature amount calculation means 103, a filter determination means 104, a multiplication means 105, and a waveform restoration means 106. are doing.

The signal processing device 100 may be partially or wholly configured by software. The signal processing device 100 may be configured, for example, by installing a program (including the signal processing program according to the embodiment) in a computer having a memory and a processor.

The first frequency analysis means 101 frequency-analyzes the first input signal x1 to obtain the first input spectrum X1.

The second frequency analysis means 102 frequency-analyzes the second input signal x2 to obtain the second input spectrum X2.

The feature amount calculating means 103 obtains a predetermined feature amount (hereinafter, referred to as “DOA feature amount F”) based on the first input spectrum X1 and the second input spectrum X2. The DOA feature amount F is a feature amount that changes according to the arrival direction of the target sound, and the details will be described later.

The feature amount calculation means 103 can obtain the DOA feature amount F by a calculation formula obtained by modifying the formula (1) or the formula (1).

In the equation (1), the first input spectrum is X ₁ , the second input spectrum is X ₂ , and the complex conjugate of the second input spectrum is X ₂ ^{* at} a certain frequency at a certain time.

The filter determining means 104 maps the DOA feature amount F with a predetermined broad-sense monotonous increasing function to obtain an emphasis filter G.

The multiplication means 105 multiplies the first input spectrum X1 by the emphasis filter G to obtain the enhancement spectrum Y.

The waveform restoration means 106 restores the signal waveform based on the enhancement spectrum Y to obtain the enhancement voice y.

Next, the DOA feature amount obtained by the feature amount calculating means 103 and the design concept of the emphasis filter obtained by the filter determining means 104 will be described.

The emphasis filter suppresses the direct sound and reflected sound of the disturbing sound coming from the second microphone M2 side (jamming sound side), and suppresses the direct sound and reflected sound from the first microphone M1 side (including the target sound side and the front direction). It is necessary to give a feature that does not suppress the direct sound and the reflected sound of the incoming target sound. Therefore, it is desired that the DOA feature amount take a large value when the sound arrives from the first microphone M1 side and a small value when the sound arrives from the second microphone M2 side. However, since the first microphone M1 side includes the front direction, such a feature is not symmetrical with respect to the arrival direction of the sound, and therefore, there is no known feature amount having the feature.

Therefore, consider a feature amount that takes a large value with respect to the front direction and a feature amount that takes a large value with respect to the second microphone M2 side. At a certain frequency at a certain time, the first input spectrum is X ₁ , the second input spectrum is X ₂ , and the complex conjugate of the second input spectrum is X ₂ ^* . For example, the equation (1-1) a feature amount _{F center} represented by), consider a feature amount _{F side} of the formula (1-2).

Here, the front direction (the direction perpendicular to the straight line connecting the positions of the two microphones) is 0 degrees, and the direction (of the second microphone M2 seen from the first microphone M1) on the second microphone M2 side is. Assuming that the spectrum of the sound source is S, the angular frequency is ω, the distance between the two microphones is d, the direction of arrival of the sound is θ (theta), and the sound velocity is c, then X ₁ and X ₂ are equations (2), respectively. And (3), and substituting equations (2) and (3) into equations (1-1) and (1-2), equations (3-1) and (3-2), respectively. The formula is obtained. The relationships between the feature quantities F _center and F _side represented by the equations (3-1) and (3-2) with respect to the sound arrival direction θ are shown in FIGS. 3 and 4, respectively.

FIG. 3 shows an example of the feature quantity F _center .

In FIG. 3, the horizontal axis is the direction of arrival of the sound source θ, and the vertical axis is the feature amount F _center . FIG. 3 is a graph showing the relationship between the arrival direction θ and the feature amount F _center when the frequency of the sound source is changed to 1 kHz, 2 kHz, and 4 kHz.

FIG. 4 shows an example of the feature amount F _side .

In FIG. 4, the horizontal axis is the direction of arrival of the sound source θ, and the vertical axis is the feature amount F _side . FIG. 4 is a graph showing the relationship between the arrival direction θ and the feature amount F _side when the frequency of the sound source is changed to 1 kHz, 2 kHz, and 4 kHz.

From FIGS. 3 and 4, the F _center has a large value with respect to the arrival direction of 0 degrees, and the F _side has a large value with respect to the second microphone M2 side with respect to 0 degrees, and the first microphone has a large value. It is a small value with respect to the M1 side.

As described above, DOA feature F, the direction of the target sound (0 ° direction; front direction) of the F _center there is a characteristic that a large value relative to the second microphone M2 side (i.e. interference sound source It can be seen that this is a feature amount obtained by using F _side, which has a characteristic of having a large value with respect to the microphone on the side of the assistant U2.

Next, the relationship between the direction of arrival of sound and the DOA feature amount will be described.

DOA features are defined by Eq. (3-3).

By substituting Eqs. (2) and (3) into Eqs. (3-3), Eqs. (4) is obtained, and by transforming Eqs., Eqs. (5) is obtained. FIG. 5 shows the relationship between the arrival direction θ of the sound represented by the equation (5) and the DOA feature amount F.

FIG. 5 is a graph showing the relationship between the sound arrival direction θ and the DOA feature amount F.

In FIG. 5, the horizontal axis is the sound source arrival direction θ, and the vertical axis is the DOA feature amount F. FIG. 5 is a graph showing the relationship between the arrival direction θ and the DOA feature amount F when the frequency of the sound source is changed to 1 kHz, 2 kHz, and 4 kHz.

From FIG. 5, the DOA feature amount F always has F = 1 in the front direction and F <1 in the second microphone M2 side (interfering sound side). On the other hand, with respect to the first microphone M1, F> 1 at a low frequency and θ close to 0 degrees at a high frequency, and F <1 at a portion close to 90 degrees at a high frequency.

As described above, the DOA feature amount F has a feature that a large value is taken when the sound arrives from the first microphone M1 side and a small value is taken when the sound arrives from the second microphone M2 side. You can see that. In other words, the DOA feature amount F has a peak in the direction from the front direction to the first microphone M1 side (the direction opposite to the disturbing sound from the assistant U2), and the second microphone M2 side from the direction in which the peak exists. It can be seen that the value becomes smaller as the direction is tilted.

The emphasis filter is obtained by mapping DOA features with a predetermined broad-sense monotonous increasing function.

FIG. 6 is a graph showing an example of an improper monotonic increase function fmap (F).

In FIG. 6, the horizontal axis is the value of the DOA feature amount F, and the vertical axis is the value of the emphasis filter G.

As can be seen from FIG. 15, the emphasis filter suppresses the sound coming from the second microphone M2 side, and does not want to suppress the sound coming from the front direction and the first microphone M1 side. Therefore, for example, the improper monotonic increase function fmap (F) is defined as in Eq. (6). FIG. 6 shows an example of fmap (F) in which the microphone spacing is 3 cm, the speed of sound is 332 m / s, and F ₀ = 0.9.

When the emphasis filter is obtained as G = fmap (F), the relationship between the sound arrival direction θ and the emphasis filter G is as shown in FIG.

FIG. 7 is a graph showing an example of an emphasis filter.

In FIG. 7, the horizontal axis is the value of the sound arrival direction θ, and the vertical axis is the value of the emphasis filter G. FIG. 7 is a graph showing the relationship between the arrival direction θ and the value of the emphasis filter G when the frequency of the sound source is changed to 1 kHz, 2 kHz, and 4 kHz.

That is, if the DOA feature amount F is larger than a value slightly smaller than 1, the emphasis filter G is set to 1, and if not, the emphasis filter G is set to be smaller than 1, so that the emphasis filter has the desired characteristics, that is, the disturbing sound. The direct sound and the reflected sound of the target sound are suppressed, but the direct sound and the disturbing sound of the target sound are not suppressed.

The emphasis filter similar to the present invention can be obtained by calculating the arrival direction θ for each frequency from, for example, the first input spectrum and the second input spectrum, but the inverse tangent function (atan, arctan, It costs a lot to calculate (written as tan ^-1 etc.). Therefore, the present invention is superior from the viewpoint of calculation cost.

(A-2) Operation of First Embodiment Next, the operation of the signal processing device 100 of the first embodiment having the above-described configuration (signal processing method of the embodiment) will be described with reference to FIG. ..

The signal processing device 100 emphasizes the target sound for the first input signal x ₁ and the second input signal x ₂ (input signal in the time domain) including the target sound source, and emphasizes the sound y (output signal in the time domain). ) Is generated.

The first frequency analysis means 101 and the second frequency analysis means are combined with the first input signal x ₁ by any frequency analysis method represented by the Fourier transform or any band division means represented by the filter bank. The second input signal x ₂ is divided into K bands, respectively, to obtain a _first input spectrum X ₁ and a second input spectrum X ₂ . Hereinafter, the first input spectrum and the second input spectrum are written as X ₁ (k) and X ₂ (k) when it is necessary to specify the band number (for example, the kth), and the band number is specified. If it is not necessary to do so, simply write X ₁ and X ₂ . The first frequency analysis means 101 gives the obtained first input spectrum X ₁ to the feature amount calculation means 103 and the multiplication means 105, and the second frequency analysis means 102 gives the obtained second input spectrum X 1. ₂ is given to the feature amount calculating means 103. The input spectrum given to the multiplication means 105 is the first input spectrum X ₁ , but the present invention is not limited to this, and the second input spectrum X ₂ may be given to the multiplication means 105. It has the same effect.

Feature calculating unit 103 has a first input spectrum _{X 1} and based on the second input spectrum _{X 2} (7) to calculate the DOA feature F by formula, giving the filter determining unit 104. The formula (7) may be used as it is for calculation, but the formula may be modified because it includes redundant calculations. Since X ₁ and X ₂ are complex numbers, rewriting them as in Eq. (8) and substituting them into Eq. (7) to obtain Eq. (9). By using the equation (9) instead of the equation (7), the number of multiplications can be reduced.

The filter determining means 104 calculates the emphasis filter G by a predetermined broad-sense monotonous increase function based on the DOA feature amount F, and gives it to the multiplication means 105. In the first embodiment, the same broadly monotonous increasing function is used for all frequencies. As a predetermined broadly defined monotonous increasing function fmap (F), for example, a function having one threshold value F ₀ defined by the equation (10) and a function having two threshold values F ₁ and F ₂ defined by the equation (11). , The sigmoid function of scale F ₃ and offset F ₀ defined by Eq. (12) can be used.

FIG. 8 is a graph showing an example of the emphasis filter G obtained by the filter determining means 104 according to the first embodiment.

8 (a), 8 (b), and 8 (c) show examples of the emphasis filter G obtained by the equations (10), (11), and (12), respectively. 8 (a), 8 (b), and 8 (c) are graphs showing the relationship between the arrival direction θ and the emphasis filter G when the frequency of the sound source is changed to 1 kHz, 2 kHz, and 4 kHz. There is. In FIGS. 8 (a), 8 (b), and 8 (c), the horizontal axis is the sound source arrival direction θ, and the vertical axis is the value of the emphasis filter G (value corresponding to the arrival direction θ).

Here, F ₀ = 0.8, F ₁ = 0.7, F ₂ = 0.9, and F ₃ = 12. There is not much difference between the equations (10) and (11) in terms of the suppression performance of the disturbing sound. On the other hand, regarding the distortion of the emphasized voice, the emphasis filter G obtained by the equation (10) tends to generate musical noise because it has only 0 or 1, but the equation (11) suppresses / suppresses it by having a transition band. Musical noise is less likely to occur because the switching of non-suppression becomes gentle. Equation (12) has characteristics that are smoother than those of equation (11), and further effects of reducing musical noise and distortion can be expected. If a slight increase in calculation cost is allowed, it is preferable to use Eq. (12).

Multiplying means 105 multiplies the enhancement filter G (emphasis gain) for each frequency to the input spectrum X _1, giving the resulting enhancement spectrum Y to the waveform restoration means 106.

The waveform restoration means 106 uses the emphasis spectrum given by the multiplication means 105 by using the waveform restoration method corresponding to the frequency analysis method or the band division method used in the first frequency analysis means 101 and the second frequency analysis means 102. The signal waveform is reconstructed based on Y, and the obtained emphasized voice y (emphasized signal) is output.

(A-3) Effect of First Embodiment According to the first embodiment, the following effects can be obtained.

In the signal processing device 100 of the first embodiment, the sound coming from the direction of the second microphone M2 side is suppressed, and the sound coming from the front direction and the direction of the first microphone M1 side is not suppressed. In the case of emphasizing the voice (target sound) of the driver U1, the target sound can be emphasized with less distortion.

In other words, the signal processing device 100 can design an emphasis filter G that suppresses the direct sound and the reflected sound of the disturbing sound at a low calculation cost, but does not suppress the direct sound and the reflected sound of the target sound. It has the effect of emphasizing the sound.

(B) Second Embodiment Hereinafter, a second embodiment of the signal processing apparatus, program and method according to the present invention will be described in detail with reference to the drawings.

(B-1) Configuration of Second Embodiment The signal processing device 200 of the second embodiment will be described as being used in the environment as shown in FIG. 2 as in the first embodiment.

Further, the internal configuration of the signal processing device 200 of the second embodiment can also be shown with reference to FIG. 1 described above.

Hereinafter, the difference between the signal processing device 200 of the second embodiment and that of the first embodiment will be described.

In the first embodiment, in the filter determining means 104, the same broadly defined monotonic increase function fmap (F) is applied to all frequencies to obtain the emphasis filter G. Therefore, as shown in FIG. 8, the characteristics of the emphasis filter G are obtained. Was different for each frequency. Especially at low frequencies, a phenomenon occurs in which the range of the arrival direction that is not suppressed becomes wide. Therefore, in the second embodiment, a broad monotonous increase function different for each frequency is applied so that the characteristics are the same at any frequency.

The configuration of the signal processing device 200 of the second embodiment is the same as the configuration of the signal processing device 100 of the first embodiment except that the filter determining means 104 is replaced with the filter determining means 204 as shown in FIG. Is.

(B-2) Operation of Second Embodiment Next, the operation of the signal processing device 200 of the second embodiment having the above configuration (signal processing method of the embodiment) will be described.

The operation of the signal processing device 200 of the second embodiment is the same as the operation of the signal processing device 100 of the first embodiment except that the operation of the filter determining means 204 is different from that of the filter determining means 104.

In the first embodiment, as shown in FIG. 7, the gain of the emphasis filter G is equal to or less than a predetermined value (for example, 0.5 or less) depending on the frequency when the arrival direction is tilted from the front direction to the side of the second microphone M2. ) (Hereinafter referred to as "cutoff arrival angle"). In other words, in the first embodiment, the range of the arrival direction that is not suppressed varies depending on the frequency. On the other hand, the filter determining means 204 of the second embodiment absorbs this variation by setting a broadly monotonous increasing function for each frequency, and cutoff arrival angles (in the arrival direction without suppression) at a plurality of frequencies. The range) is getting closer. The filter determining means 204 is not limited to a method of obtaining a broad monotonous increase function for each frequency that suppresses variation in the cutoff arrival angle for each frequency (variation in the range of the arrival direction that is not suppressed). For example, any processing can be applied.

The filter determining means 204 calculates the emphasis filter G by a predetermined broad-sense monotonous increasing function based on the DOA feature amount F, and gives it to the multiplication means 105. In the second embodiment, a broad monotonous increase function that differs for each frequency is used. Here, the DOA feature of the k-th frequency is written as F (k), and the emphasis gain of the k-th frequency is written as G (k). Let _{fk be the} k-th frequency, and define a function that converts the arrival direction and frequency number into DOA features by Eq. (13). Then, as the broadly defined monotonous increasing function fmap _k (F (k)) of the predetermined k-th frequency, for example, one arrival direction threshold θ ₀ defined by the equation (14) is defined by the function or the equation (15). that two DOA threshold theta _1, function with the theta _2, it is possible to use the scale F _a, sigmoid function of the offset arrival direction theta ₀ as defined in (16).

FIG. 9 is a graph showing an example of the emphasis filter G obtained by the filter determining means 204 according to the second embodiment.

9 (a), 9 (b), and 9 (c) show examples of the emphasis filter G obtained by the equations (14), (15), and (16), respectively. 9 (a), 9 (b), and 9 (c) are graphs showing the relationship between the arrival direction θ and the emphasis filter G when the frequency of the sound source is changed to 1 kHz, 2 kHz, and 4 kHz. There is. In FIGS. 9A, 9B, and 9C, the horizontal axis is the sound source arrival direction θ, and the vertical axis is the value of the emphasis filter G (value corresponding to the arrival direction θ).

Here, θ ₀ = 15, θ ₁ = 20, θ ₂ = 10, and _Fa = 12. In the emphasis filter G (FIG. 8) of the first embodiment, the range of the non-suppressing arrival direction changed for each frequency, but in the emphasis filter G (FIG. 9) of the second embodiment, the high frequency first. Except for the microphone M1 side of No. 1, the range of the arrival direction that is not suppressed does not change even if the frequency changes. Although the characteristics of Eq. (16) change depending on the frequency, the cutoff arrival angle at which the gain of the emphasis filter G is 0.5 is constant regardless of the frequency. That is, if the emphasis gain is calculated using the equations (13) to (16), the number of times the second microphone M2 side, that is, the disturbing sound side (passenger seat side) is suppressed is common to all frequencies. Can be set directly.

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

In the signal processing device 200 of the second embodiment, the range of the arrival direction in which the emphasis gain is not suppressed can be given in the same manner at all frequencies, and the range can be set by the angle of the arrival direction itself, which is more appropriate. The effect is that the target sound can be emphasized with less distortion.

(C) Third Embodiment Hereinafter, a third embodiment of the signal processing apparatus, program and method according to the present invention will be described in detail with reference to the drawings.

(C-1) Configuration of Third Embodiment The signal processing device 200 of the third embodiment will be described as being used in the environment as shown in FIG. 2 as in the first and second embodiments. ..

Further, the internal configuration of the signal processing device 300 of the third embodiment can also be shown with reference to FIG. 1 described above.

Hereinafter, the difference between the signal processing device 200 of the third embodiment and that of the second embodiment will be described.

In the second embodiment, the number of times the direction of arrival on the disturbing sound side is suppressed is set for all frequencies. However, it is difficult to make the microphone interval sufficiently wide with respect to the signal wavelength at low frequencies (the wavelength of 100 Hz is about 3.3 m, but the microphone interval in the car is generally several cm). Information on the arrival direction (DOA feature amount in the present invention) obtained by numerical calculation at a low frequency is generally ambiguous (the estimation error becomes large in the sense of the arrival direction estimation). Therefore, in the third embodiment, the range of the arrival direction in which the emphasis gain is not suppressed is widened (widened to the side of the second microphone M2; an interfering sound is emitted) at a frequency lower than a predetermined frequency (for example, a frequency band of 250 Hz or less). It is designed to spread to the side of the assistant U2).

The configuration of the signal processing device 300 of the third embodiment is the same as the configuration of the signal processing device 100 of the first embodiment except that the filter determining means 104 is replaced with the filter determining means 304.

(C-2) Operation of Third Embodiment Next, the operation of the signal processing device 300 of the third embodiment having the above configuration (signal processing method of the embodiment) will be described.

The operation of the signal processing device 300 of the third embodiment is the same as the operation of the signal processing device 100 of the first embodiment except that the operation of the filter determining means 304 is different from that of the filter determining means 104.

The filter determining means 304 calculates the emphasis filter G by a predetermined broad-sense monotonous increase function based on the DOA feature amount F, and gives it to the multiplication means 105. In the second embodiment, a broad monotonous increase function that differs for each frequency is used. Here, the DOA feature of the k-th frequency is written as F (k), and the emphasis gain of the k-th frequency is written as G (k). Letting the kth frequency be f _k , the function for converting the arrival direction and the frequency number into the DOA feature is defined by the equation (17). Then, as the broadly defined monotonic increasing function fmap _k (F (k)) of the predetermined k-th frequency, for example, a function having one arrival direction threshold value θ ₀ defined by the equation (18) can be used.

FIG. 10 is a graph showing a comparison between the emphasis filter G obtained by the filter determination means 204 according to the second embodiment and the emphasis filter G obtained by the filter determination means 304 according to the third embodiment.

FIG. 10A shows the emphasis filter G obtained by using the above equations (13) and (14) in the filter determining means 204 in the second embodiment. Further, FIG. 10B shows an example of the emphasis filter G obtained by the equations (17) and (18) in the filter determining means 304 according to the third embodiment. 10 (a) and 10 (b) are graphs showing the relationship between the arrival direction θ and the emphasis filter G when the frequency of the sound source is changed to 1 kHz, 2 kHz, and 4 kHz. In FIGS. 10A and 10B, the horizontal axis is the sound source arrival direction θ, and the vertical axis is the value of the emphasis filter G (value corresponding to the arrival direction θ).

Here, θ ₀ = 5 and F ₀ = 0.97. Than 10, the function for converting the number of the arrival direction and frequency in DOA feature quantity Φ the upper limit of (phi) that has a F _0, it can be confirmed that the suppression was not DOA range of low frequencies is wider ..

(C-3) Effect of Third Embodiment According to the third embodiment, the effect can be added to the effects of the first and second embodiments.

In the signal processing device 300 of the third embodiment, at a low frequency where the information about the arrival direction obtained by the numerical calculation is ambiguous, a wide range of the arrival direction that is not suppressed can be secured, so that the distortion of the target sound at the low frequency is distorted. It is reduced and has the effect of emphasizing the target sound with less distortion.

(D) Fourth Embodiment Hereinafter, a fourth embodiment of the signal processing apparatus, program and method according to the present invention will be described in detail with reference to the drawings.

(D-1) Configuration of Fourth Embodiment The signal processing device 400 of the fourth embodiment will be described as being used in the environment as shown in FIG. 2 as in the first to third embodiments. ..

Further, the internal configuration of the signal processing device 400 of the fourth embodiment can also be shown with reference to FIG. 1 described above.

Hereinafter, the difference between the signal processing device 400 of the fourth embodiment and the first to third embodiments will be described.

In the first to third embodiments, the emphasis filter G that suppresses only the passenger side (passenger U2 side) is assumed in the case where the two microphones M1 and M2 are set in front of the driver U1 in the automobile A. Designed. On the other hand, in the fourth embodiment, the emphasis filter that emphasizes (does not suppress) only the front direction by using the DOA feature amount in the present invention is applied.

As shown in FIG. 1, the configuration of the signal processing device 400 of the fourth embodiment is the first except that the feature amount calculating means 103 and the filter determining means 104 are replaced with the feature amount calculating means 403 and the filter determining means 404, respectively. It is the same as the configuration of the signal processing device 100 of the first embodiment.

(D-2) Operation of Fourth Embodiment Next, the operation of the signal processing device 400 of the fourth embodiment having the above configuration (signal processing method of the embodiment) will be described.

Next, the operation of the signal processing device 400 of the fourth embodiment having the above-described configuration will be described. The operation of the signal processing device 400 of the fourth embodiment is that of the first embodiment, except that the operations of the feature amount calculating means 403 and the filter determining means 304 are different from those of the feature amount calculating means 103 and the filter determining means 104. The operation is the same as that of the signal processing device 100.

The feature amount calculation means 403 calculates two DOA feature amounts F and F'by the equation (19) based on the first input spectrum X ₁ and the second input spectrum X _2, and gives them to the filter determination means 404. .. When the two DOA features are arranged with respect to the sound arrival direction θ, it becomes Eq. (20).

FIG. 11 is a graph showing the relationship between the sound arrival direction θ and the DOA feature amount F'in equation (20).

FIG. 11 is a graph showing the relationship between the arrival direction θ and the DOA feature amount F'when the frequency of the sound source is changed to 1 kHz, 2 kHz, and 4 kHz. In FIG. 11, the horizontal axis is the sound source arrival direction θ, and the vertical axis is the DOA feature amount F'.

Looking at FIG. 11, it can be confirmed that the left and right sides are exactly reversed from the DOA feature amount F (FIG. 5).

The filter determining means 404 calculates the emphasis filter G by a predetermined broad-sense monotonous increasing function based on the two DOA features F and F', and gives it to the multiplication means 105. The predetermined broadly defined monotonous increasing function includes fmap (F) according to the first embodiment, Φ (φ, k) and fmap _k (F (k)) according to the second embodiment, and the third embodiment. Either Φ (φ, F ₀ , k) or fmap _k (F (k)) may be used, but here, as an example, a predetermined broad-sense monotonic increase function of the second embodiment will be used. .. In the fourth embodiment, the emphasis filter G is calculated using the equation (21).

FIG. 12 is an explanatory diagram showing an example of the emphasis filter G obtained by the filter determining means 404 according to the fourth embodiment.

12 (a), 12 (b), and 12 (c) are obtained when the formulas (14), (15), and (16) are used as the fmap _k (F (k)), respectively. An example of the emphasis filter G is shown. 12 (a), 12 (b), and 12 (c) are graphs showing the relationship between the arrival direction θ and the emphasis filter G when the frequency of the sound source is changed to 1 kHz, 2 kHz, and 4 kHz. There is. In FIGS. 12 (a), 12 (b), and 12 (c), the horizontal axis is the sound source arrival direction θ, and the vertical axis is the value of the emphasis filter G (value corresponding to the arrival direction θ).

Here, θ ₀ = 20, θ ₁ = 15, θ ₂ = 25, and _Fa = 12. From FIG. 12, it can be seen that an emphasis filter that emphasizes (does not suppress) only the front direction is obtained.

(D-3) Effect of Fourth Embodiment According to the fourth embodiment, the following effects can be obtained as compared with the first to third embodiments.

In the signal processing device 400 of the fourth embodiment, it is possible to achieve a unique effect that the target sound can be emphasized by limiting the range of the arrival direction that the emphasis filter G (emphasis gain) does not suppress to the front direction.

(E) Other Embodiments The present invention is not limited to each of the above embodiments, and modified embodiments as illustrated below can also be mentioned.

(E-1) In the above embodiment, the signal processing device is described as restoring the waveform of the emphasized spectrum Y and outputting the emphasized voice y, but outputs the enhanced spectrum Y without restoring the waveform. Is also good. Further, both the emphasized spectrum Y and the emphasized voice y may be output. In that case, the waveform restoration means 106 may be excluded.

100 ... Signal processing device, 101 ... First frequency analysis means, 102 ... Second frequency analysis means, 103 ... Feature amount calculation means, 104 ... Filter determination means, 105 ... Multiplying means, 106 ... Waveform restoration means, M1 ... First microphone (first sound collecting device), M2 ... Second microphone (second sound collecting device).

Claims (10)

A first frequency analysis means for obtaining a first input spectrum by frequency analysis of a first input signal input from the first sound picking device, and
A second frequency analysis means for obtaining a second input spectrum by frequency analysis of the second input signal input from the second sound pickup device, and
Based on the first input spectrum obtained by the first frequency analysis means and the second input spectrum obtained by the second frequency analysis means, the position of the first sound collecting device and the second The values in the front direction and the direction on the first sound collecting device side are set larger than the front direction perpendicular to the straight line connecting the positions of the sound collecting devices, and the values in the direction on the second sound collecting device side are taken. The feature amount calculation means for calculating the first feature amount, and
A filter determining means for obtaining an emphasis filter by mapping the first feature amount calculated by the feature amount calculating means with a predetermined broad-sense monotonous increasing function.
A signal processing apparatus comprising: a multiplication means for obtaining an emphasis spectrum by multiplying a first input spectrum obtained by the first frequency analysis means by a enhancement filter obtained by the filter determination means. The first feature amount is on the side of the first sound collecting device with respect to the front direction perpendicular to the straight line connecting the position of the first sound collecting device and the position of the second sound collecting device. The signal processing device according to claim 1, wherein a peak exists in the direction, and the value becomes smaller as the direction is inclined from the direction of the peak toward the second sound collecting device. The feature amount calculating means has a second feature amount that has a large value with respect to the front direction and a third feature amount that has a large value with respect to the direction toward the second sound collecting device than the front direction. The signal processing apparatus according to claim 1 or 2, wherein the first feature amount is calculated using the above. The method according to any one of claims 1 to 3, wherein the filter determining means maps the first feature quantity using a broad-sense monotonous increase function different for each frequency, and obtains the emphasis filter for each frequency. Signal processing device. The signal processing apparatus according to claim 4, wherein the filter determining means sets a broad monotonous increasing function of each frequency so as to match the range of the arrival direction that is not suppressed in the emphasis filter for each frequency. .. The filter determining means sets a broad monotonous increasing function in the emphasis filter in a low frequency band below a predetermined frequency to widen the range of the arrival direction that is not suppressed as compared with a high frequency band higher than the predetermined frequency. The signal processing apparatus according to claim 4. The filter determining means sets a broad monotonous increasing function in the emphasis filter in the low frequency band to widen the range of the arrival direction that is not suppressed as compared with the high frequency band toward the second sound pickup device. The signal processing apparatus according to claim 6. The signal according to any one of claims 1 to 3, wherein the filter determining means uses the first feature amount to set a broadly defined monotonous increasing function for obtaining an emphasis filter that emphasizes only the front direction. Processing equipment. Computer,
A first frequency analysis means for obtaining a first input spectrum by frequency analysis of a first input signal input from the first sound picking device, and
A second frequency analysis means for obtaining a second input spectrum by frequency analysis of the second input signal input from the second sound pickup device, and
Based on the first input spectrum obtained by the first frequency analysis means and the second input spectrum obtained by the second frequency analysis means, the position of the first sound collecting device and the second The values in the front direction and the direction on the first sound collecting device side are set larger than the front direction perpendicular to the straight line connecting the positions of the sound collecting devices, and the values in the direction on the second sound collecting device side are taken. The feature amount calculation means for calculating the first feature amount, and
A filter determining means for obtaining an emphasis filter by mapping the first feature amount calculated by the feature amount calculating means with a predetermined broad-sense monotonous increasing function.
A multiplication means for obtaining an emphasis spectrum by multiplying the first input spectrum obtained by the first frequency analysis means by the enhancement filter obtained by the filter determination means.
A signal processing program characterized by inputting an emphasis spectrum obtained by the multiplication means and restoring a signal waveform to function as a waveform restoration means for obtaining an emphasized voice. In the signal processing method
It has a first frequency analysis means, a second frequency analysis means, a feature amount calculation means, a filter determination means, and a multiplication means.
The first frequency analysis means frequency-analyzes the first input signal input from the first sound collecting device to obtain a first input spectrum.
The second frequency analysis means frequency-analyzes the second input signal input from the second sound collecting device to obtain a second input spectrum.
The feature amount calculation means is based on the first input spectrum obtained by the first frequency analysis means and the second input spectrum obtained by the second frequency analysis means, and the first sound collecting device is used. The values in the front direction and the direction on the side of the first sound collecting device are set larger than those in the front direction perpendicular to the straight line connecting the position of the second sound collecting device and the position of the second sound collecting device. Calculate the first feature amount that keeps the value in the direction of the sound device small,
The filter determining means maps the first feature amount calculated by the feature amount calculating means with a predetermined broad-sense monotonous increase function to obtain an emphasis filter.
The multiplication means is a signal processing method, characterized in that a first input spectrum obtained by the first frequency analysis means is multiplied by an emphasis filter obtained by the filter determination means to obtain an emphasis spectrum.

Images