JP5060438B2 - Sound collector - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a sound collecting device capable of suppressing noise. <P>SOLUTION: The sound collecting device includes: a sound receiving section 1 for outputting a voice signal before processing of intensity according to sound pressure of input sound, a time differential signal of intensity according to a time differential value of the sound pressure, and x-differential signal and y-differential signal according to a space differential value of the sound pressure in each axis direction of a two-dimensional orthogonal coordinate system; and a voice processing section 2 for generating a voice signal after processing, in which noise from a position other than a predetermined target position is suppressed, by load summing of the voice signal before processing, the time differential signal, the x-differential signal and the y-differential signal, which are output by the sound collection section 1. The voice processing section 2 suitably updates a reference distance which is used for determination of the load used for generation of the voice signal after processing. Compared to a case that the reference distance is always matched with the distance to the target position, sound (noise) from a position other than the target position is suppressed. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to a sound collection device.

  2. Description of the Related Art Conventionally, there has been provided a sound collection device that generates a post-processing audio signal in which sound from a predetermined target position is selectively reflected using a pre-processing audio signal output from a microphone or the like (for example, Patent Documents). 1-3). As the target position, a position where the sound source (speaker or the like) of the target sound is considered to be present is selected.

In addition, the pre-processed sound signal having an intensity corresponding to the sound pressure of the input sound, the time differential signal having an intensity corresponding to the time differential value of the sound pressure, and the sound pressure in each axial direction of the two-dimensional orthogonal coordinate system. In a spatiotemporal gradient method using sound receiving means for outputting x differential signals and y differential signals according to spatial differential values, the weight sum of pre-processed audio signals, temporal differential signals, x differential signals, and y differential signals is calculated. A technique of generating an audio signal after processing is known (for example, see Non-Patent Document 1). After the processing, which is an integral value of the square of the intensity of the processed audio signal over a predetermined load sum calculation time, the load used for the above load sum is subjected to a constraint condition that the gain with respect to the sound from the target position is constant. If the MV (Minimum Variance) method that minimizes the power is used, the sound after the target position is suppressed and the sound from the target position is selectively reflected in the processed audio signal obtained. Is done.
JP 2006-154314 A JP 2000-47699 A Japanese Patent No. 3484112 Ono, Ando, “Measurement of sound field and directivity control, 22nd Sensing Forum, pp. 305-310, 2005.

  In this type of sound collecting device, it is desirable to suppress as much as possible ambient noise and reverberant sound (hereinafter referred to as “noise”) from positions other than the target position in the processed audio signal. For example, when the processed speech signal output from the sound collection device is used for speech recognition, noise reduction in the processed speech signal leads to an improvement in recognition accuracy.

  This invention is made | formed in view of the said reason, The objective is to provide the sound collection device which can suppress a noise more.

  According to the first aspect of the present invention, the pre-processed speech signal having the intensity corresponding to the sound pressure of the input sound, the time differential signal having the intensity corresponding to the time differential value of the sound pressure, and each axial direction of the two-dimensional orthogonal coordinate system. A sound receiving unit that outputs an x differential signal and a y differential signal corresponding to a spatial differential value of the sound pressure, a pre-processing voice signal, a time differential signal, an x differential signal, and a y differential signal output by the sound receiving unit, respectively. And a sound processing unit that generates a processed sound signal based on a load sum of the load and the sound processing unit is configured to determine a load used for the load sum according to a predetermined target position where a target sound is considered to exist. MV that minimizes the processed power that is the integral value of the square of the intensity of the processed audio signal over a predetermined load sum calculation time under the constraint that the gain for the sound from the reference position is constant. The reference position is determined by law. There is a possibility that the reference distance, which is the distance between the sound receiving unit and the reference position, is updated so that the power after processing is minimized, which is selected from a straight line passing through the target position starting from the sound receiving unit. The update operation is performed periodically.

  According to the present invention, it is possible to further suppress noise from other than the target position in the processed audio signal as compared with the case where the reference position used for determining the load is always matched with the target position.

  According to a second aspect of the present invention, in the first aspect of the present invention, a pre-processing power calculation unit that calculates a pre-processing power that is an integral value of the square of the intensity of the pre-processing voice signal over the load sum calculation time, and a pre-processing power A determination unit that compares the pre-processing power obtained by the calculation unit with a predetermined volume threshold value, and the audio processing unit is the update operation, if the pre-processing power is less than the volume threshold value as a result of the comparison in the determination unit, The reference distance is not changed.

  According to the present invention, it is possible to avoid changing the reference distance when the sound pressure (volume) of the sound incident on the sound receiving unit is insufficient.

  According to a third aspect of the present invention, in the first or second aspect of the present invention, the pre-processing power calculation unit that calculates a pre-processing power that is an integral value of the square of the intensity of the pre-processing voice signal over the load sum calculation time. Calculating a processing ratio, which is a ratio of the post-processing voice power output by the post-processing power calculation unit to the post-processing power calculation unit that calculates the post-processing power and the pre-processing voice power output by the pre-processing power calculation unit; A determination unit that compares the obtained processing ratio with a predetermined target sound threshold, and the sound processing unit does not change the reference distance if the processing ratio obtained by the determination unit is equal to or greater than the target sound threshold during the update operation. It is characterized by that.

  According to the present invention, it is possible to avoid changing the reference distance when there is little noise.

  According to a fourth aspect of the present invention, in the first or second aspect of the present invention, the difference from the target position is determined by the sum of the weights of the pre-processing audio signal, the time differential signal, the x differential signal, and the y differential signal output by the sound receiving unit. A dead processing unit that generates a dead voice signal that does not reflect voice, a pre-processing power calculation unit that calculates a pre-processing power that is an integral value of the square of the strength of the pre-processing voice signal over the load sum calculation time, and the load Power after desensitization processing for calculating power after desensitization, which is an integral value of the square of the intensity of the dead sound signal over the sum calculation time, and power after desensitization for the pre-process voice power output by the pre-process power calculation unit And a determination unit that calculates the dead processing ratio that is a ratio of the voice power after the dead processing output by the calculation unit and compares the obtained dead processing ratio with a predetermined noise threshold, and the voice processing unit, during the update operation, Size Insensitive treatment ratio obtained in part is characterized in that it does not change the reference distance is less than the noise threshold.

  According to the present invention, it is possible to avoid changing the reference distance when there is little noise.

  According to a fifth aspect of the present invention, in the third aspect of the invention, when the processing ratio obtained by the determination section is less than the target sound threshold during the update operation, the voice processing section minimizes the reference distance and the processing ratio. It is characterized by updating to a value such as

  The invention of claim 6 is the invention of claim 4, further comprising a post-processing power calculation unit that calculates post-processing power, and the voice processing unit has an insensitive processing ratio obtained by the determination unit during the update operation equal to or greater than a noise threshold value. The reference distance is updated to a value that minimizes the ratio of the post-processing power output from the post-processing power calculation unit to the post-insensitive processing power output from the post-insensitive processing power calculation unit. Features.

  The invention according to claim 7 is the invention according to any one of claims 1 to 6, wherein the sound processing unit outputs the pre-processing sound signal, the time differential signal, and the x differential signal output by the sound receiving unit when updating the reference distance. Based on the y differential signal, it is determined whether or not the number of sound sources is one, and if the number of sound sources is one, the distance to the sound source is estimated, The distance is used as the updated reference distance.

  The invention of claim 8 is characterized in that, in the invention of any one of claims 5 to 7, the speech processing unit sets the change width of the reference distance in one update operation to a predetermined upper limit width or less.

  According to the present invention, the distortion of the processed audio signal due to the large change of the reference distance in one update operation is suppressed, and the uncomfortable feeling given by the sound obtained by converting (reproducing) the processed audio signal is reduced. .

  The invention according to claim 9 is the invention according to any one of claims 5 to 7, wherein the operation that the voice processing unit may perform with respect to the reference distance during the update operation does not change the reference distance and increases the reference distance by a predetermined unit. The reference distance is increased by an amount, or the reference distance is decreased by a predetermined unit decrease amount.

  According to the present invention, the distortion of the processed audio signal due to the large change of the reference distance in one update operation is suppressed, and the uncomfortable feeling given by the sound obtained by converting (reproducing) the processed audio signal is reduced. .

  According to the first aspect of the present invention, the sound processing unit has a constant gain for a sound from a reference position determined according to a predetermined target position at which a target sound is regarded as a load used for the load sum. Is determined by an MV method that minimizes the processed power, which is the integral value of the square of the intensity of the processed audio signal over a predetermined load sum calculation time. The position is selected from a straight line that passes through the target position starting from the sound receiving unit, and the reference distance, which is the distance between the sound receiving unit and the reference position, can be updated so that the post-processing power is minimized. Since the renewal operation is performed periodically, it is possible to further suppress noise from other than the target position in the processed audio signal, compared to the case where the reference position used for determining the load is always matched with the target position. .

  According to the second aspect of the present invention, the pre-processing power calculation unit that calculates the pre-processing power that is an integral value of the square of the intensity of the pre-processing voice signal over the load sum calculation time, and the pre-processing power calculation unit are obtained. A determination unit that compares the pre-process power with a predetermined volume threshold value, and the audio processing unit does not change the reference distance when the pre-process power is less than the volume threshold value as a result of the comparison in the determination unit during the update operation. Therefore, it is possible to avoid changing the reference distance when the sound pressure (volume) of the sound incident on the sound receiving unit is insufficient.

  According to the invention of claim 3, the pre-processing power calculation unit that calculates the pre-processing power that is an integral value of the square of the intensity of the pre-processing audio signal over the load sum calculation time, and the post-processing that calculates the post-processing power The processing ratio obtained by calculating the processing ratio, which is the ratio of the post-processing voice power output by the post-processing power calculation section to the pre-processing voice power output by the power calculation section and the pre-processing power calculation section, and the obtained processing ratio for a predetermined purpose A determination unit that compares with a sound threshold, and the sound processing unit does not change the reference distance if the processing ratio obtained by the determination unit is equal to or greater than the target sound threshold during the update operation. Can be avoided.

  According to the fourth aspect of the present invention, a dead sound signal that does not reflect the sound from the target position is generated by the load sum of the pre-processing sound signal, the time differential signal, the x differential signal, and the y differential signal output from the sound receiving unit. A desensitization processing unit, a pre-processing power calculation unit that calculates pre-processing power that is an integral value of the square of the intensity of the pre-processing audio signal over the load sum calculation time, and the intensity of the insensitive audio signal over the load sum calculation time Insensitive post-insensitive power computation unit that computes power after insensitive processing, which is an integral value of the square of, and post-insensitive speech output from the post-insensitive power computing unit with respect to the pre-processing audio power output from the pre-processing power computing unit And a determination unit that calculates the insensitive processing ratio that is a power ratio and compares the obtained insensitive processing ratio with a predetermined noise threshold, and the voice processing unit has the insensitive processing ratio obtained by the determining unit during the update operation. Noise Does not change the reference distance is less than the value, it is possible to prevent the reference distance is changed when the noise is small.

  According to the invention of claim 8, the voice processing unit sets the change width of the reference distance in one update operation to a predetermined upper limit width or less, and according to the invention of claim 9, the voice processing unit is in the update operation. The operations that may be performed with respect to the reference distance are either not changing the reference distance, increasing the reference distance by a predetermined unit increase amount, or decreasing the reference distance by a predetermined unit decrease amount. Distortion of the processed audio signal due to a large change in the reference distance in one update operation is suppressed, and the uncomfortable feeling given by the sound obtained by converting (reproducing) the processed audio signal is reduced.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

  In this embodiment, as shown in FIG. 1, a sound from a predetermined target position is selectively reflected using a sound receiving unit 1 that converts input sound into an electrical signal and an output of the sound receiving unit 1. And an audio processing unit 2 that generates the processed audio signal.

Sound receiving unit 1, a voice signal pre-processing of intensity corresponding to the sound pressure f of the input sound, and time differential signal of intensity corresponding to the time derivative f _t of the sound pressure f, predetermined in the x-axis direction the sound and x differential signal corresponding to the spatial differential value f _x of pressure f, y differential signal corresponding to the spatial differential value f _y of the sound pressure f for a given y-axis direction orthogonal to the x-axis direction of the And output respectively.

Specifically, the sound receiving unit 1 is provided with a square apex arrangement as shown in FIG. 2, for example, and generates an original audio signal having intensities f _{A to} f _D according to the sound pressures of the input sounds. Spatio-temporal gradient processing unit that generates four microphones 10A to 10D and a pre-processed audio signal, a time differential signal, an x differential signal, and a y differential signal using the original audio signal output from each of the microphones 10A to 10D ( (Not shown). The microphones 10 </ b> A to 10 </ b> D and the spatiotemporal gradient processing unit as described above can be realized by a known technique, and thus detailed illustration is omitted.

That is, the spatiotemporal gradient processing unit uses the intensity f _{A to} f _D of the output of each of the microphones 10A to 10D and the length d of one side of the above square to calculate the sound pressure f and the time differential value f _t . , spatial differential value f _x, and f _y, obtained by the following equation (1) to (4).

f = (f _A + f _B + f _C + f _D ) / 4 (1)
f _t = df / dt (2)
f _x = {(f _A + f _B ) − (f _C + f _D )} / 2d (3)
f _y = {(f _A + f _C ) − (f _B + f _D )} / 2d (4)
The sound processing unit 2 selects a sound from a predetermined reference position by taking a weighted sum of the above-mentioned unprocessed sound signal output from the sound receiving unit 1, the time differential signal, and the x differential signal and the y differential signal. A post-processed audio signal that is automatically reflected is generated. That is, the intensity of the processed speech signal, the sound pressure f and the time differential value f _t and the spatial differential value f _x, the intensity corresponding to the weighted sum of the f _y. Since the audio processing unit 2 can be realized by a known electronic circuit, a detailed circuit diagram and the like are omitted.

Here, if the input vector F is defined as F = (f f _t f _x f _y ) ^T and the load vector W is defined as W = (w w _t w _x w _y ) ^T , The intensity is expressed as W ^H F.

A known MV (Minimum Variance) method is used to determine the load vector W used for the load sum. The MV method is a method for controlling directivity by using the spatiotemporal gradient method, and the spatiotemporal gradient method has been proposed as one of the methods for determining the optical flow that is the apparent velocity field in the moving image. (See Reference 1). More specifically, in the MV method, the gain in the processed audio signal is made constant with respect to the audio from the reference position, and the square of the intensity of the processed audio signal over the predetermined period is averaged. The optimization with constraints is performed to minimize the value (dispersion) E [(W ^H F) ² ]. Thereby, the sound from other than the reference position is suppressed in the processed audio signal.

Hereinafter, a method for determining the load vector W will be specifically described (see References 2 to 4). For simplicity, assume that there is one sound source and a coordinate system having the position of the sound receiving unit 1 (hereinafter referred to as “observation point”) as the origin as shown in FIG. The speed of sound is c, the coordinates of the sound source are (x, y, z), the distance between the sound source and the observation point is r = (x ² + y ² + z ² ) ^1/2 , and the sound field formed by the position of the sound source is If g is set, the sound field formed at the observation point, that is, the sound pressure f is expressed by the following equation.

X sound pressure f at an observation point, the spatial derivative of y-direction _(gradient) f x, _{f y} is

here,
ξ _x = x / r ² , ξ _y = y / r ² (8)
Is called the intensity gradient,
τ _x = x / cr, τ _y = y / cr (9)
Is called the time gradient in the x and y directions.

  Equations (6) and (7) showing the spatial gradient in the x and y directions of the sound pressure f (t) at the observation point are expressed as a vector R = (-x, -y, -z) (ie, from the sound source to the observation point) Rewrite using | R | = r)

It becomes. Then _{f (t), f t (} t), when ∇f (t) is observed, these weighted sum is

It is expressed. Here, w and w _t are real constants, and W _S = (w _x , w _y , 0) is a unit vector. Substituting equation (10) into equation (11),

It becomes. Therefore, the load sum of the spatiotemporal gradient is expressed as the sum of filters having different directivity characteristics H (R) and H _t (R) for f (t) and f _t (t), respectively. When H (R) = α, equation (13) is

And can be transformed. Here, generally, if the angle formed by the two vectors a and b is θ, the following formula holds.

Using the formula of equation (18), equation (16) can be rewritten as

Here, from | W _S | = 1,

It is expressed by the sphere equation. In the case of w + α = 0, the formula (15) is R · W _S = 0 (22)
It becomes. When H _t (R) = α, equation (14) becomes

Therefore, if the angle formed by the vector R and W _S is θ (R), | W _S | = 1

It becomes. Therefore, equation (23) becomes

It becomes.

From the equations (21), (22), and (25), the following properties are obtained for H (R) and H _t (R).
1) The two directivity characteristics H (R) and H _t (R) have a rotationally symmetric body with Ws as the axis. 2) When H (R) = 0, the distribution of R has a diameter 1 / w (w ≠ 0). ) To form a spherical surface or plane (w = 0) 3) When H _t (R) = 0, the distribution of R is a conical surface or plane (w _t = 0) with apex angle 2cw _t (w _t ≠ 0). 4) The intersection of the distributions of R when H (R) = 0 and H _t (R) = 0 forms a circle or a plane. The above directivity characteristics H (R) and H _t (R) are set to H ₁ ( R ₁ ) and H ₂ (R ₁ ) are defined as follows. Where R ₁ is the position vector of the reference position, r ₁ = | R ₁ | is the distance from the observation point to the reference position (hereinafter referred to as “reference distance”), and n _1x and n _1y are the vector R ₁ and The x component and y component of the unit vector R ₁ / r _{1 in} the same direction, and n _1x ² + n _1y ² = 1.

Further, two constraint conditions such as the following formulas (28) and (29) are set for these H ₁ (R ₁ ) and H ₂ (R ₁ ).
W ^H H ₁ (R ₁ ) ＝ p… (28)
W ^H H ₂ (R ₁ ) ＝ q… (29)
Here, p and q are positive real constants, respectively. Then, with respect to sound from a sound source reference position indicated by the vector R _1, the gain due to the load sum with weight vector W is constant p + jωq, in order to compensate for this, in FIG. 2 In the speech processing unit 2, a first-order low-pass filter 22 of (p + jωq) ⁻¹ is provided after the load sum calculation unit 21 that takes the load sum. As a result, the gain of the entire audio processing unit 2 is 1 with respect to the sound from the sound source at the reference position indicated by the vector R ₁ . In this embodiment, p = 1 / r ₁ and q = 1 / c.

Then, the weight vector W that most suppresses the sound (noise) from other than the reference position without suppressing the sound from the reference position in the processed audio signal is obtained under the conditions of equations (28) and (29). , The power of the processed speech signal in the observation time interval Γ (hereinafter referred to as “processed power”) P _c

Is obtained by using a Minimum Variance Beamformer (MV method) for minimizing. That is, the observation window Γ corresponds to the load sum calculation time in the claims. This solution is expressed as the following equations (31) and (32).

However, Bij (i, j = a, x, y) in Expression (32) is expressed by Expression (33), and b _a (t), b _x (t), b in Expression (33) _y (t) is represented by the equations (34) to (36), respectively.

Here, according to the inventor's simulation, even when the reference position S is on the z-axis as shown in FIGS. 2 and 3A and 3B, as shown in FIGS. Even when the reference position S is a position deviating from the z-axis, it is on a straight line connecting the observation point (that is, the position of the sound receiving unit 1 and the origin in the above coordinate system) O and the reference position S. Sound from the sound source is hardly suppressed by the processed audio signal. In other words, even if the reference position S does not coincide with the target position where the sound source desired to be selectively left in the processed audio signal is assumed to exist, the reference position S is on the straight line connecting the observation point O and the target position. If there is a position S, the goal of avoiding sound suppression from the sound source at the target position is achieved. In other words, the reference distance r ₁ does not necessarily match the distance between the observation point and the target position (hereinafter referred to as “target distance”), and the reference distance r ₁ and the target distance are different from each other. However, the sound from the sound source at the target position is not suppressed in the processed audio signal. In each simulation shown in FIGS. 2 to 5, the number of sound sources is one. FIGS. 3A, 3B and 5 are diagrams showing the relationship between the sound suppression amount (hereinafter referred to as “noise suppression amount”) in the processed audio signal and the position of the sound source. The darker the position, the higher the amount of noise suppression for the sound from that sound source when it is assumed that the sound source exists at that position. In FIGS. 3A and 5, the reference distance r ₁ is 0.5 m, and in FIG. 3B, the reference distance r ₁ is 1 m.

Further, according to the simulation of the present inventor, a point on the spherical surface (white line in FIG. 5) having the observation point O as the center and the reference distance r ₁ as the radius, and other than the reference position S is comparatively relatively. The noise suppression amount in the processed speech signal in the case where the noise source is at a position deviating from the straight line connecting the observation point and the reference position such that the noise suppression amount increases at the time depends on the reference distance r ₁ . . For example, consider a case where the reference position S is on the z-axis. As shown in FIG. 6, when the angle φ formed by the noise source N with respect to the z axis is 30 ° and is at a distance of 1.0 m, as shown in FIGS. 7 and 8, the reference distance r _{When 1} is 0.5 m, the noise suppression amount is 14.5 dB, whereas when the reference distance r ₁ is 1 m, which is the same as the distance to the noise source N, the noise suppression amount is 25. It is 6 dB. 8A shows the waveform of the unprocessed audio signal, FIG. 8B shows the waveform of the processed audio signal when the reference distance r ₁ is 0.5 m, and FIG. 8C shows the waveform of FIG. the reference distance r ₁ shows the waveform of the processed audio signal in case of a 1.0 m. Further, as shown in FIG. 9, there are two noise sources N1 and N2 in a direction where the angle φ formed with respect to the z-axis is 45 °, and the distance between one noise source N1 and the observation point is 1.0 m. Even when the two noise sources N1 and N2 are separated from each other such that the distance between the other noise source N2 and the observation point is 0.5 m, the reference distance r ₁ is set as shown in FIG. When the distance from the observation point to the middle between the two noise sources N1 and N2, the noise suppression amount has a peak value. That is, if appropriately changing the reference distance r _1, compared with the case of always match the target position a reference position, there is a possibility that the sound (noise) is more suppressed from the non-target position.

Therefore, in this embodiment, as the sound from other than the target position (noise) is suppressed more, at any time updates the reference distance r ₁ used to determine the load vector W.

In detail, the voice processor 2, at predetermined time intervals, and performs an update operation for updating the weight vector W, it is possible to update the reference distance r ₁ at this time. Specific contents of the update operation will be described with reference to the flowchart of FIG.

When the sound processing unit 2 starts the update operation (S1), it first determines whether or not the sound volume input to the sound receiving unit 1 is sufficient (S2). That is, in the present embodiment, the power of the pre-processing audio signal in the observation window Γ determined with reference to the start time of the update operation (hereinafter referred to as “pre-processing power”) P _f .

And a determination unit 30 that compares the obtained pre-process power P _f with a predetermined silence threshold. When the determination unit 30 determines that the pre-processing power P _f is less than the silence threshold, the sound processing unit 2 determines that the input volume is not sufficient and is not suitable for updating the reference distance r ₁ and the load vector W. The update operation is terminated (S3). On the other hand, when the determination unit 30 determines that the pre-processing power P _f is equal to or greater than the silence threshold, the audio processing unit 2 determines that the volume is sufficient, and _first, by the MV method using the reference distance r ₁ before the update. The load vector W is determined (S4).

After determining the load vector W, the voice processing unit 2 determines whether or not there is noise (S5). That is, in the present embodiment, the post-processing power calculation unit 32 that calculates the power (post-processing power) _Pc of the post-processing audio signal obtained by using the load vector W determined in step S4 by the equation (30) is provided. The determination unit 30 has a ratio of the post-processing power P _c output from the post-processing power calculation unit 32 to the pre-processing power P _f output from the pre-processing power calculation unit 31 (hereinafter referred to as “processing ratio”). And compare P _c / P _f with a predetermined target sound threshold. If the determination unit 30 determines that the processing ratio is less than the target sound threshold, that is, if the sound pressure as a whole of the processed sound signal is reduced to some extent with respect to the sound pressure f of the unprocessed sound signal, The voice processing unit 2 determines that noise is present, and if the determination unit 30 determines that the processing ratio is equal to or higher than the target sound threshold, the voice processing unit 2 determines that no noise exists. On the other hand, if the determination unit 30 determines that there is no noise because it is determined that the processing ratio is less than the target sound threshold, there is no point in updating the reference distance, so the speech processing unit 2 proceeds directly to step S3. The update operation is terminated. On the other hand, when it is determined that there is noise, the voice processing unit 2 determines whether or not the number of sound sources is one (S6). The target sound threshold is a positive constant smaller than 1. Since the determination unit 30, the pre-processing power calculation unit 31, and the post-processing power calculation unit 32 can be realized by well-known electronic circuits, detailed illustration is omitted.

  A specific method for the determination in step S6 will be described. The speech processing unit 2 uses the covariance matrix S estimated from the observation window Γ.

The number of sound sources is estimated by calculating the rank rank (S). Such estimation of the number of sound sources is described in Reference Document 4. That is, if rank rank (S) is 2, the number of sound sources is 1, if rank rank (S) is 3, the number of sound sources is 2, and if rank rank (S) is 4, the number of sound sources is The number is 3 or more.

When it is determined in step 6 that the number of sound sources is one, the sound processing unit 2 calculates the distance to the sound source and updates the reference distance r ₁ using the obtained distance as a new reference distance. , and update operation again load vector W in accordance with the new reference distance r ₁ (S7), and terminates the subsequent update operation proceeds to step S3. A specific method for calculating the distance to the sound source will be described (see Reference 5). First, τ _x , τ _y , ξ _x , ξ _y in equations (8) and (9) are obtained by the method of least squares. The evaluation function in the short observation window Γ
J = ∫ _Γ {(f _x + ξ _x f + τ _x f _t ) ² + (f _y + ξ _y f + τ _y f _t ) ² } dt… (39)
And When the equation (39) is partially differentiated with respect to τ _x , τ _y , ξ _x , ξ _y and set to 0, the following equation is obtained.

Using the matrix elements of the covariance matrix S of equation (21), equations (40) and (41) are

Rewritten. By solving the equations (42) and (43), τ _x , τ _y, ξ _{x and} ξ _y are obtained as follows.

  The distance r to the sound source can be obtained by applying the least square method from the equations (8) and (9). Evaluation function

And if this is partially differentiated by 1 / r and set to 0

It becomes. Solving this

The distance r to the sound source is obtained as follows. In step S7, the audio processing unit 2, the formula (38) (44) (45) a distance r obtained in (48), it is taken as the new reference distance r _1.

If it is determined in step S6 that there are two or more sound sources, the audio processing unit 2 maximizes the noise suppression amount (that is, minimizes the post-processing power P _c and the processing ratio, respectively). explore the Do) reference distance r ₁ (S8), and updates the values obtained in the reference distance r ₁ step S8, update operation again load vector W in accordance with the new reference distance r ₁ (S9 After that, the process proceeds to step S3 to end the update operation. Specifically, the operation of determining the load vector W after increasing / decreasing the reference distance r ₁ by a predetermined unit width and obtaining and storing the processed power _Pc from the processed power calculation unit 32 is repeated to increase / decrease the load vector W. Regardless of the direction, the reference distance r ₁ is updated so that the post-processing power P _c increases (that is, the noise suppression amount decreases).

  According to the above configuration, it is possible to suppress noise more than when the reference distance is always set as the target distance.

  In addition, since the method is based on the spatiotemporal gradient method, the sound receiving unit 1 can be downsized compared to a general beamformer or superdirective microphone, and all calculations are performed in the time domain. Compared to other methods that require computation in the frequency domain, the computation cost can be reduced.

As shown in FIG. 12, a dead sound processing unit 33 that generates a dead voice signal in which a sound from the target position is not reflected (that is, the target position is a dead point), and the power of the dead voice signal (hereinafter referred to as “dead process”). In addition to providing a post-insensitive power calculation unit 34 that calculates _Pz ), in step S5, the determination unit 30 determines the post-insensitive power P _z output by the post-insensitive power calculation unit 34. The ratio (hereinafter referred to as “insensitive processing ratio”) P _Z / P _f to the pre-processing power P _f output by the pre-processing power calculation unit 31 may be compared with a predetermined noise threshold. In this case, if the determination unit 30 determines that the insensitive processing ratio is equal to or greater than the noise threshold, there is some sound from other than the target position, so the sound processing unit 2 determines that noise is present and vice versa. If the determination unit 30 determines that the processing ratio is less than the noise threshold, the speech processing unit 2 determines that there is no noise. The noise threshold is a positive constant smaller than 1. Since the dead processing unit 33 and the post-dead processing power calculation unit 34 can be realized by a well-known electronic circuit like the voice processing unit 2 and the like, detailed circuit diagrams and the like are omitted. The operation of the dead processing unit 33 will be described in detail. For H ₁ (R ₁ ) and H ₂ (R ₁ ) defined in Expressions (26) and (27), the vector R ₁ indicating the reference position is replaced with the vector R ₀ indicating the target position, and the following expression Two constraint conditions such as (49) and (50) are set.
W _Z ^H H ₁ (R ₀ ) = 0 (49)
W _Z ^H H ₂ (R ₀ ) = 0 (50)
Load vector satisfying this condition (hereinafter referred to as “dead point load vector”) W _Z = (w ′ w ′ _t w ′ _x w ′ _y ) Insensitivity obtained by the load sum W _Z ^H F (t) by ^T The audio signal forms a dead point at the target position. The power _Pz after dead processing is expressed by the following equation.

Further, in the example of FIG. 12, the post-processing power calculation unit 32 is added, and in step S8, the post-processing power calculation unit 32 outputs the desensitized power _Pz output by the post-dead processing power calculation unit 34. the ratio P _{C /} P _Z of the processed power P _C may be searched reference distance r ₁ that minimizes.

Further, in step S7 and step S9, the upper limit for one variation width of the reference distance r ₁ per update operation (hereinafter, referred to as "upper limit width".) Is provided, the criteria obtained in step S7 or step S8 distance and the absolute value of the difference between the reference distance r ₁ before update may be changed reference distance r ₁ by the amount of the upper limit width when the value exceeds the upper limit width. By this configuration, the reference distance distortion of the processed speech signal due to the r ₁ is changed greatly suppressed, discomfort voice converting the processed audio signal (reproduction) provides is reduced.

Alternatively, in step S7 and step S9, the fluctuation range of the reference distance r1 in _one update operation may be constant. Specifically, for example, the audio processing unit 2 does not change the reference distance r _1, increasing the reference distance r ₁ by the unit increment, decreases the reference distance r ₁ by the unit decline, the three kinds of operation the inner shell, the reference distance r ₁ after the update operation is completed, to perform one of the operations such as closest to the reference distance obtained by the search operation or step S8 in step S7. In this case, the unit increase width and the unit decrease width may be different from each other. If this configuration is adopted, distortion of the processed audio signal due to a large change in the reference distance is suppressed, and the uncomfortable feeling given by the sound obtained by converting (reproducing) the processed audio signal is reduced.
<List of references>
Reference 1: Shigeru Ando “Velocity vector distribution measurement system using spatio-temporal differential calculation of images” Transactions of the Society of Instrument and Control Engineers 22-12, 1330/1336 (1986)
Reference 2: N. Ono, T. Arita, Y. Senjo, and S. Ando, “Directivity steeringprinciple for biomimicry silicon microphone”, Proc. Int. Conf. Solid State Sensors,
Actuators, and Microsystems (Transducers'05), pp. 792-795, 2005.
Reference 3: Ono, Ando, “Measurement of sound field and directivity control, 22nd Sensing Forum document, pp. 305-310, 2005.
Reference 4: Ono, Arita, Chiaki, Ando, “Theory of Directional Control and Sound Source Separation Based on Spatiotemporal Gradient Measurement, Proc. Of the Spring Meeting of the Acoustical Society of Japan 2005, 2-6-13, pp. 607-608, 2005.
Reference 5: Shigeru Ando, Hiroyuki Shinoda, Katsuya Ogawa, Satoshi Mitsuyama "3D sound source localization sensor system based on spatiotemporal gradient method" Vol. 29, No. 5, p520-528, 1993

It is a block diagram which shows embodiment of this invention. It is explanatory drawing which shows an example of the positional relationship of the reference | standard position and sound receiving part in the same as the above. (A) (b) is explanatory drawing which shows an example of distribution of the noise suppression amount in the case of FIG. 2, respectively (a) shows the case where a reference distance is 0.5 m, (b) shows a reference distance. The case of 1.0 m is shown. It is explanatory drawing which shows another example of the positional relationship of the reference | standard position and sound receiving part in the same as the above. FIG. 5 is an explanatory diagram showing an example of a noise suppression amount distribution in the case of FIG. 4. It is explanatory drawing which shows an example of the positional relationship of the position of a noise source and sound receiving part in the same as the above. It is explanatory drawing which shows the relationship between the reference distance in the case of FIG. 6, and the amount of noise suppression. (A)-(c) is explanatory drawing which respectively shows operation | movement same as the above, (a) shows the waveform of the audio signal before a process, (b) is the audio signal after a process in case a reference distance is 0.5 m. (C) shows the waveform of the processed audio signal when the reference distance is 1.0 m. It is explanatory drawing which shows another example of the positional relationship of the position of a noise source and sound receiving part in the same as the above. FIG. 10 is an explanatory diagram showing a relationship between a reference distance and a noise suppression amount in the case of FIG. 9. It is a flowchart which shows the update operation | movement in the same as the above. It is a block diagram which shows another form same as the above.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Sound receiving part 2 Audio | voice processing part 30 Judgment part 31 Pre-processing power calculation part 32 Post-processing power calculation part 33 Dead process part 34 Power calculation part after dead process

Claims (9)

A pre-processed speech signal having an intensity corresponding to the sound pressure of the input sound, a time differential signal having an intensity corresponding to the time differential value of the sound pressure, and a spatial differential of the sound pressure in each axial direction of a two-dimensional orthogonal coordinate system A sound receiving unit that outputs an x differential signal and a y differential signal corresponding to the value;
An audio processing unit that generates a processed audio signal by a weighted sum of the pre-processing audio signal, the time differential signal, the x differential signal, and the y differential signal output by the sound receiving unit;
The voice processing unit, under the constraint that the gain for the sound from the reference position determined according to a predetermined target position where the target voice is considered to be the load used for the load sum is constant, It is determined by the MV method that minimizes the processed power that is the integral value of the square of the intensity of the processed audio signal over a predetermined load sum calculation time,
The reference position is selected from a straight line passing through the target position starting from the sound receiving unit,
A sound collection device that periodically performs an update operation that may update a reference distance, which is a distance between a sound receiving unit and a reference position, so that power after processing is minimized.   An unprocessed power calculator that calculates an unprocessed power that is an integral value of the square of the intensity of the unprocessed audio signal over the load sum calculation time; and the unprocessed power obtained by the unprocessed power calculator is a predetermined volume threshold value. 2. The determination unit according to claim 1, wherein the audio processing unit does not change the reference distance during the update operation if the pre-processing power is less than a volume threshold value as a result of the comparison in the determination unit. Sound collecting device.   A pre-processing power calculation unit that calculates a pre-processing power that is an integral value of the square of the intensity of the pre-processing voice signal over the load sum calculation time, a post-processing power calculation unit that calculates post-processing power, and a pre-processing power calculation A determination unit that calculates a processing ratio that is a ratio of the post-processing audio power output by the post-processing power calculation unit to the pre-processing audio power output by the unit and compares the obtained processing ratio with a predetermined target sound threshold value; 3. The sound collection device according to claim 1, wherein the sound processing unit does not change the reference distance during the update operation if the processing ratio obtained by the determination unit is equal to or greater than a target sound threshold.   A dead processing unit that generates a dead voice signal that does not reflect a voice from a target position based on a load sum of a pre-processing voice signal, a time differential signal, an x differential signal, and a y differential signal output by the sound receiving unit; An unprocessed power calculation unit that calculates an unprocessed power that is an integral value of the square of the intensity of the unprocessed audio signal over the calculation time, and an insensitivity that is an integral value of the square of the intensity of the insensitive audio signal over the load sum calculation time. A desensitization processing ratio, which is a ratio of the desensitized power output unit output by the desensitized power calculation unit to the unprocessed power calculation unit that calculates the post-processing power and the unprocessed power calculation unit output by the pre-processing power calculation unit. And a determination unit that compares the insensitive processing ratio obtained with the calculation with a predetermined noise threshold, and the speech processing unit has a reference distance if the insensitive processing ratio obtained by the determination unit is less than the noise threshold during the update operation. Sound pickup apparatus according to claim 1 or claim 2, wherein it does not change.   The voice processing unit updates the reference distance to a value that minimizes the processing ratio when the processing ratio obtained by the determination unit is less than the target sound threshold during the update operation. Item 4. The sound collecting device according to Item 3. Provided with a post-processing power calculator that calculates post-processing power,
When the dead processing ratio obtained by the determination unit is greater than or equal to the noise threshold during the update operation, the speech processing unit uses the post-processing power relative to the post-dead processing power output by the post-dead power calculation unit. 5. The sound collecting device according to claim 4, wherein the sound collecting device is updated to a value that minimizes the ratio of the post-processing power output by the arithmetic unit.   When updating the reference distance, the sound processing unit determines whether or not the number of sound sources is one based on the unprocessed sound signal, the time differential signal, the x differential signal, and the y differential signal output from the sound receiving unit. In addition, when the number of sound sources is one, the distance to the sound source is estimated, and the estimated distance to the sound source is used as the updated reference distance. The sound collecting device according to any one of the above.   The sound processing device according to claim 5, wherein the sound processing unit sets a change width of the reference distance in one update operation to a predetermined upper limit width or less.   The operations that the voice processing unit may perform with respect to the reference distance during the update operation include either not changing the reference distance, increasing the reference distance by a predetermined unit increase amount, or decreasing the reference distance by a predetermined unit decrease amount. The sound collecting device according to claim 5, wherein the sound collecting device is a sound collecting device.