JP2012037429A - Method for creating filter for beam-forming device and apparatus for creating filter for beam-forming device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for creating a filter for a beam-forming device with which a sound source generation position can be highly accurately detected even in a low frequency.SOLUTION: The method includes: a transfer function calculation step (S504) for calculating transfer functions from a plurality of virtual sound source positions to a microphone on the basis of a distance from the virtual sound source position to the microphone and a frequency of a sound source; and a filter group creation step (S504) for creating a filter group obtained by configuring a filter in such a manner that a value becomes greater than 0, in which the value is obtained by subtracting either a second transfer function sum, which is obtained by arithmetically processing and mutually adding each of the transfer functions from the virtual sound source positions in a non-target direction to the microphone through a filter, or an absolute value of the second transfer function sum from a first transfer function sum, which is obtained by arithmetically processing and mutually adding the transfer functions from the virtual sound source positions in a target direction to the microphone through the filter so as to match phases with each other.

Description

  The present invention relates to a method for producing a filter for a beam forming apparatus and an apparatus for producing a filter for a beam forming apparatus.

  Conventionally, the generation direction (position) of an unknown sound source generated in a vehicle or facility is detected.

  Delay sum beamforming, which is a conventional technique, is a method of detecting the direction of a sound source using a microphone array composed of a plurality of microphones. Delay sum beamforming enhances a signal in a certain direction by matching the phases of sound waves detected by each microphone and summing them, and detects the more emphasized direction as the direction of the sound source (Patent Document 1).

JP 2004-279390 A

  However, when the frequency of sound waves generated from the sound source is low, the wavelength of the sound wave becomes longer relative to the size of the microphone array, and the phase difference of the sound waves detected by each microphone becomes small. There is a problem that accuracy decreases.

  The present invention has been made to solve such a problem, and can detect a sound source generation position with high accuracy even if a sound wave to be detected has a low frequency.

  In order to solve the above problems, a method for creating a filter for a beamforming apparatus according to the present invention includes a plurality of filter groups, and each of detection signals is subjected to arithmetic processing by a filter group having different sound source directions and estimated target directions. This is a method of creating a filter that can obtain the directional characteristics of sound waves. The method for creating a filter for a beam forming apparatus according to the present invention includes a transfer function calculation stage and a filter group creation stage. The filter group emphasizes the detection signal in the target direction by adding together the detection signals calculated by the plurality of filters to which the detection signals from the plurality of microphones installed in the space are respectively input. In the transfer function calculation stage, each transfer function from the plurality of virtual sound source positions to the microphone is calculated based on the distance from the plurality of virtual sound source positions where the sound source can exist to the microphone and the frequency of the sound source. The filter group creation stage creates a filter group in which the filter is configured such that the value obtained by subtracting either the second transfer function sum or the absolute value of the second transfer function sum from the first transfer function sum is greater than zero. To do. The first transfer function sum is obtained by performing arithmetic processing on the respective transfer functions from the virtual sound source position in the target direction to the microphone among the plurality of virtual sound source positions and adding them to each other so as to match each other's phase. The second transfer function sum is obtained by calculating each transfer function from the virtual sound source position in the direction other than the target direction to the microphone among the plurality of virtual sound source positions and adding them to each other.

  In addition, the beam forming apparatus filter creation apparatus according to the present invention includes a plurality of filter groups, and the detection signals are calculated and processed by the filter groups having different target directions and the estimated direction of the sound source, so that the directional characteristics of the sound waves are obtained. This is a filter creation device obtained. The filter forming apparatus for a beam forming apparatus according to the present invention includes a transfer function calculating unit and a filter group generating unit. The filter group emphasizes the detection signal in the target direction by adding together the detection signals calculated by the plurality of filters to which the detection signals from the plurality of microphones installed in the space are respectively input. The transfer function calculation means calculates each transfer function from the plurality of virtual sound source positions to the microphone based on the distance from the plurality of virtual sound source positions where the sound source can exist to the microphone and the frequency of the sound source. The filter group creation means creates a filter group in which the filters are configured such that a value obtained by subtracting either the second transfer function sum or the absolute value of the second transfer function sum from the first transfer function sum is greater than zero. To do. The first transfer function sum is obtained by performing arithmetic processing on the respective transfer functions from the virtual sound source position in the target direction to the microphone among the plurality of virtual sound source positions and adding them to each other so as to match each other's phase. The second transfer function sum is obtained by calculating each transfer function from the virtual sound source position in the direction other than the target direction to the microphone among the plurality of virtual sound source positions and adding them to each other.

  A filter is created in which the sum of the transfer functions in the direction other than the target direction or the sum of the transfer functions in the directions other than the target direction or the absolute value thereof is greater than zero. As a result, the width of the main lobe of the beam can be narrowed, and a peak can always be provided in the target direction, and the width of the beam and the size of the side lobe can be optimized. As a result, the detection of the sound source generation position can be realized with high accuracy even if the sound wave to be detected has a low frequency.

It is explanatory drawing for demonstrating the creation method of the filter for beam forming apparatuses which concerns on 1st Embodiment of this invention, and the principle of a preparation apparatus. It is a figure which shows the example of arrangement | positioning of the several microphone which comprises a microphone array. It is a block diagram of the preparation apparatus of the filter for beam forming apparatuses which concerns on 1st Embodiment of this invention. It is a block diagram of the preparation apparatus of the filter for beam forming apparatuses which concerns on 2nd Embodiment of this invention. It is a figure which shows the flowchart of the preparation method of the filter for beam forming apparatuses which concerns on 2nd Embodiment of this invention. It is a figure which shows the evaluation result of the beam forming characteristic by the beam forming apparatus using the filter produced with the production method of the filter for beam forming apparatuses which concerns on 2nd Embodiment of this invention, and the conventional beam forming apparatus.

Hereinafter, a method for creating a filter for a beamforming device and a device for creating a filter for a beamforming device according to an embodiment of the present invention will be described in detail with reference to the drawings.
(First embodiment)
FIG. 1 is an explanatory diagram for explaining the principle of a filter forming method for a beam forming apparatus according to the first embodiment of the present invention and the generating apparatus.

  In the method of creating a filter for a beam forming apparatus according to the present embodiment, each of the positions from the position (virtual sound source position) where the sound source X (ω) (ω is an angular frequency) can exist to each microphone 20a constituting the microphone array 20 is provided. The transfer function H (ω) is calculated. Then, the transfer function H (ω) by calculation is used in the creation of the filter 10a. As will be described later, the transfer function H (ω) can be calculated based on the distance from the virtual sound source position to each microphone 20a and the frequency of the sound wave.

  The filter 10a calculates each transfer function H (ω) in each microphone 20a. The filter group 10 includes a filter 10a provided for each microphone 20a, and adds the transfer functions H (ω) calculated by the filters 10a to each other. The filter group 10 emphasizes the value of the transfer function H (ω) regarding the direction (target direction) estimated as the direction of the sound source among the directions of the virtual sound source with respect to the microphone array 20. Here, the virtual sound source is a virtual sound source at a virtual sound source position.

  A plurality of filters 10 having different target directions are created. Thereby, at the time of beam forming, it is possible to detect the direction of the sound source in question by calculating the sound waves from the sound source in question with the filter group 10 having different target directions and comparing the magnitude of the output Y (ω). it can. That is, by creating a plurality of filter groups 10 having different target directions, it is possible to specify the direction in which the sound waves from the sound source in question are more emphasized, thereby obtaining the directional characteristics of the sound waves.

  In the present embodiment, the filter group 10 is created so as to match the phases of the transfer functions H (ω) from the target direction. Furthermore, the sum of the transfer functions H (ω) from the target direction in phase with each other and the sum of the transfer functions in directions other than the target direction (hereinafter referred to as “non-target direction”) or the absolute value thereof is subtracted. The filter group 10 is created so as to be larger than zero.

  FIG. 2 is a diagram illustrating an arrangement example of a plurality of microphones 20 a constituting the microphone array 20.

  As shown in FIG. 2, the microphone array 20 can be composed of 32 microphones 20a. However, the interval and quantity of the microphones 20a can be optimized from the viewpoint of the detection sensitivity of the direction of the sound source 1 based on the estimated frequency, sound volume, target range, and the like of the sound source 1.

  The method and apparatus for creating a filter for a beam forming apparatus according to the present embodiment will be described in more detail in comparison with a conventional method for creating a filter for a beam forming apparatus.

As shown in FIG. 1, the characteristics of the sound source 1 at the azimuth angle θ and the elevation angle φ from the microphone array 20 are X (ω), and the transfer function from the sound source 1 to each microphone 20a at the distance r is H _{r, θ, φ. , N} (ω) (where n = 1, 2,..., N (N is the number of microphones forming a microphone array)). _Let the transfer characteristic of the filter be G _n (ω). Then, from the ratio between the characteristic X (ω) of the sound source 1 and the output signal Y (ω) added and output by the adder, the directivity characteristic D (θ, φ) of this processing system is expressed by the following equation (1). ).

Here, if the sound source 1 is a virtual sound source, the transfer function H _{r, θ, φ, n} (ω) is obtained from the distance r from the virtual sound source to each microphone 20a and the frequency f (= ω / 2π) of the sound source 1. Can be calculated. That is, assuming that the sound wave generated from the virtual sound source is a sine wave, the amplitude of the transfer function H _{r, θ, φ, n} (ω) is inversely proportional to the distance r, and the phase p (= ωr / C, where C is the speed of sound. The transfer function H _{r, θ, φ, n} (ω) can be calculated in consideration of being proportional to the distance r (formula (1)).

  Further, the discretized direction in the target range is represented by the subscript m (m = 1, 2,..., M), and the directional characteristic vector d, transfer function matrix H, Consider a filter vector g. These are given by the following equation (2) (T is a transposed matrix).

  Here, the target range is a range where the virtual sound source position exists.

  When Expression (2) is used, the directivity vector d can be simply expressed by the following Expression (3).

  Moreover, Formula (3) can be transformed into Formula (4) (+ is a pseudo inverse matrix).

  The filter group 10 for performing beam forming can be created by solving Equation (4).

  As shown in Expression (2) and Expression (4), the filter 10a is created so that the filter group 10 including the N filters 10a has directivity in the target direction. Then, the number (M) of directions in which the filter groups 10 having different target directions are discretized in the target range is created.

  At the time of beam forming, a sound wave signal detected by the microphone array 20 is digital Fourier transformed into a frequency domain, calculated by each filter 10a, and gain is obtained by calculating the sum of the calculation results. The obtained gain in each direction m is displayed on the contour map. That is, a set of gains for each direction is displayed as a beam on the contour map. Thereby, the direction with the largest gain, that is, the direction of the apex of the beam can be detected as the direction of the sound source 1.

The delay sum beamforming filter vector g, which is a general beamforming method, includes only arithmetic elements that delay the transfer function _{Hm, n} in each direction m so that the phases of sound waves from the target direction coincide with each other. That is, the delay sum beamforming uses the relationship between the sound wave phase p caused by the difference in distance r from the sound source 1 to each microphone 20a of the microphone array 20 and the position of the sound source 1, and the sound wave phase from the target direction. The filter 10a is configured so that sound waves from the target direction are emphasized by so-called synchronous addition in which they are added together.

  However, in the delay sum beamforming, when the frequency of the sound wave generated from the sound source 1 is low, the wavelength of the sound wave becomes relatively long with respect to the size of the microphone array array 20 (that is, the interval between the microphones 10a), and each microphone 20a. Since the phase difference of the sound wave detected by the method becomes small, there is a problem that the detection accuracy of the direction of the sound source decreases.

  To solve such a problem, a method of setting the sum of transfer functions from non-target directions to 0 can be considered.

  Considering m = 1 as the target direction, m = 2, 3,..., M as the non-target direction, and setting the non-target direction element among the elements of the directivity vector d to 0, the filter vector g is (5).

  The filter created by the expression (5) forces the transfer function sum from the non-target direction out of the sum (hereinafter referred to as “transfer function sum”) obtained by processing the transfer functions by the filter 10a and adding them to each other. Set to zero. This method is called a so-called NULL method.

  According to the filter according to Expression (5), the width of the main lobe of the beam displayed on the contour map can be narrowed even when the sound wave generated by the sound source 1 has a low frequency, thereby improving the search accuracy. Can do. However, the beam forming using the filter according to the equation (5) has a problem that the side lobe of the beam becomes large because the transfer function sum of the directivity vector d in the non-target direction is forced to be zero.

  FIG. 3 is a block diagram of the filter forming apparatus for the beam forming apparatus according to the present embodiment.

  The filter forming apparatus 3 for a beam forming apparatus according to the present embodiment can be configured by a computer and a program. However, part or all of the filter forming apparatus 3 for the beam forming apparatus can be configured by an electronic circuit.

  The beam forming apparatus filter creation device 3 according to the present embodiment includes a transfer function calculation unit (transfer function calculation unit) 300 and a filter group generation unit (filter group generation unit) 301. The filter forming device 3 for the beam forming device receives the target range, distance r, and frequency f of the sound source from the input device of the computer, generates a filter group 10 by calculation, and outputs it from the output device of the computer. Can do.

  The transfer function calculation unit 300 receives the target range, the distance r, and the sound source frequency f, and calculates the transfer function H (ω) with respect to all virtual sound source positions in the target range by Expression (1). The target range can be, for example, an elevation angle of −30 degrees to 30 degrees and an azimuth angle of −45 degrees to 45 degrees at a point where the radius is 1 m with respect to the center of the microphone array 20. The virtual sound source can be set every 3 degrees in the target range.

  The filter group creation unit 301 receives the input of the transfer function H (ω) calculated by the transfer function calculation unit 300 and calculates the filter group 10 having the virtual sound source position as the target direction for all virtual sound source positions in the target range by calculation. Create and output.

  The filter group creation unit 301 creates the filter group 10 as follows. That is, the phase of each transfer function H (ω) related to the target direction is matched, and the value obtained by subtracting the transfer function sum of the non-target direction or its absolute value from the sum of the transfer functions from the target direction with the phase matched is 0. The filter group 10 is created so as to increase.

  The following expression (6) is an expression for the filter forming apparatus for the beam forming apparatus according to the present embodiment to create the filter group 10.

Here, h (ω) is a transfer function in the target direction, and h _d (ω, d1) to h _d (ω, d8) is a subtraction of the transfer function sum in the non-target directions d1 to d8 from the transfer function sum in the target direction. (That is, the difference between the transfer function sum in the target direction and the transfer function sum in the non-target direction).

  In Expression (6), the transfer function sum in the target direction (first transfer function sum) is set to 1. This means that the filter 10a is created so as to match the phase of each transfer function from the virtual sound source position in the target direction to the microphone 20a.

Filter group 10 beamforming device according to the present embodiment, the difference h _d of the transfer function the sum of the target direction transfer function sum and non-target direction (the second transfer function sum) becomes a predetermined value in each of the microphone arrays 20 Create as follows. Here, the difference between the transfer function sum in the target direction and the transfer function sum in the non-target direction is 1 or less.

Here, instead of the difference h _d of the transfer function sum of transfer functions sum and non-target direction of the target direction to create to a predetermined value a1 to a8, the transfer function sum of the target direction (the formula (6) The difference between 1) and the absolute value of the transfer function sum in the non-target direction may be set to a predetermined value. Here, the difference between the transfer function sum in the target direction and the absolute value of the transfer function sum in the non-target direction is 1 or less.

The difference between the transfer function sum in the target direction and the transfer function sum in the non-target direction or the absolute value thereof (hereinafter referred to as “additional transfer function difference h _d ”) Set to be smaller than the transfer function sum. That is, the addition transfer function difference _hd is set to be larger than zero. Adding transfer function difference _{h d} can be set to, for example, 0.5. The addition transfer function difference h _d is the corresponding non-target direction may be a larger value as the distance between the target direction.

The non-target directions d1 to d8 are preferably set in the vicinity of the target direction. By suppressing the transfer function sum of non-target direction as adding transfer function difference h _d is greater than a predetermined value, it is possible to narrow the width of the beam displayed contour map. Furthermore, a peak can always be provided in the target direction. Therefore, even if the sound wave to be detected is a low frequency, the sound source generation position can be detected with high accuracy.

  Furthermore, by setting the addition transfer function difference to an appropriate value, the beam width can be narrowed and the beam side lobe can be optimized.

Specifically, the expression (6) can be expressed as the following expression (7). Here, the non-target direction d1~d8 2-9 in the formula (6), adding the transfer function difference h _d are expressed replaced each expression specific transfer function.

It is possible to produce a beam forming apparatus according to this embodiment by creating a filter group 10 seeking filter vector G ₁ ~G _N by solving the simultaneous equations shown in Equation (7).

The method for producing a filter for a beam forming apparatus and the apparatus for producing a filter for a beam forming apparatus according to the present embodiment have the following effects.
-The width of the main lobe of the beam can be narrowed, and a peak can always be provided in the target direction, and the width of the beam and the size of the side lobe can be optimized. As a result, the detection of the sound source generation position can be realized with high accuracy even if the sound wave to be detected has a low frequency.
(Second Embodiment)
A method for producing a filter for a beam forming apparatus and a filter producing apparatus for a beam forming apparatus according to a second embodiment of the present invention will be described in detail. In addition, the description which overlaps with description of 1st Embodiment is abbreviate | omitted.

  The difference between the embodiment and the first embodiment is as follows. That is, the first embodiment satisfies the condition that the difference between the transfer function sum in the target direction and the transfer function sum in the non-target direction of the signal detected by the microphone array 20 becomes a predetermined value in the creation of the filter group 10. Create a filter. On the other hand, in the present embodiment, a condition for making the transfer function sum 0 for any of the non-target directions is added in the creation of the filter. In addition, the filter group 10 monitors the beam obtained by the calculation obtained by the calculation process of the transfer function by the created filter group 10 and reduces the maximum value of the side lobe level of the beam (hereinafter referred to as “maximum side lobe level”). Adjust.

  FIG. 4 is a block diagram of an apparatus for creating a filter for a beamforming apparatus according to the present embodiment.

A filter forming apparatus 3 for a beam forming apparatus according to the present embodiment includes a transfer function calculation unit 300, a filter group generation unit (filter group generation unit, filter adjustment unit) 301, a beam calculation unit (beam calculation unit) 302, and random number generation. Part (filter adjusting means) 303. Creating apparatus 3 of filter beamforming device receives coverage, the distance r, the sound source frequency f, the initial value of the maximum sidelobe level SLL, the initial value of the addition the transfer function difference h _d, the maximum number of loops, the input, A filter group 10 is generated by calculation and output.

  The filter group generation unit 301 receives the input of the transfer function H (ω) calculated by the transfer function calculation unit 300, and performs an operation on the filter group 10 having the virtual sound source position as the target direction for all virtual sound source positions in the target range. Create and output.

  The filter group creation unit 301 creates the filter group 10 as follows. The value obtained by adjusting the phase of each transfer function H (ω) from the target direction and subtracting the transfer function sum in the non-target direction or its absolute value from the transfer function sum from the target direction after adjusting the phase is greater than zero. The filter group 10 is created so as to satisfy (first condition). Further, the filter group 10 is created so that the transfer function sum in other non-target directions becomes 0 (second condition). At this time, the filter group generation unit 301 can cause the random number generation unit 301 to generate a random number and determine the other non-target direction based on the random number generated by the random number generation unit 303.

In addition, the filter group creation unit 301 transmits the generated filter group 10 to the beam calculation unit 302, and causes the beam calculation unit 302 to calculate a beam using the transfer function H (ω). Then, the beam calculated by the beam calculation unit 402 is monitored, and the filter group 10 is adjusted so that the maximum side lobe level SLL of the beam decreases. Adjustment of the filter group 10, or a transfer function sum on the basis of the random number update by regenerated random number generating unit 303 resets the non-target direction to 0, the initial value of the addition the transfer function difference h _d set It can be done by changing. The adjustment of the filter group may be performed, for example, within the number of times less than the input maximum number of loops.

  The beam calculation unit 302 receives the filter group 10 from the filter group generation unit 301 and receives the transfer function H (ω) from the transfer function calculation unit 300. Then, the beam is calculated as follows. That is, the target direction of the filter group 10 generated by the filter group generation unit 301 is fixed to (0, 0), and the transfer function H (ω) from all virtual sound source positions in the target range to each microphone 20a is the target direction. The calculation is performed by the filter group 10 which is (0, 0). Then, the absolute value of the transfer function sum after the arithmetic processing is calculated for each direction of the virtual sound source position. The calculated beam can be obtained as a set of absolute values of the transfer function sum calculated in this way.

  The random number generation unit 303 generates a random number for determining a non-target direction in which the transfer function sum is zero.

  The following formula (8) is a formula for creating the filter for the beam forming apparatus filter creation method and creation apparatus according to the present embodiment.

Here, h (ω) is a transfer function in the target direction, and h _d (ω, d1) to h _d (ω, d8) are added transfer function differences. Further, h (ω, sm1) to h (ω, smm) are transfer functions in a direction to which a NULL point is applied (that is, the transfer function sum is set to 0) among non-target directions.

  Also in Expression (8), the transfer function sum in the target direction is set to 1 as in the first embodiment. This means that the filter 10a is created so as to match the phase of each transfer function from the virtual sound source position in the target direction to the microphone 20a.

Also in this embodiment, like the first embodiment, in the creation of filter group 10, it sets the sum transfer function difference h _d a predetermined value a1 to a8. Furthermore, in this embodiment, the filter is created so that the transfer function sum is 0 (a NULL point is applied) in at least one of the non-target directions.

The direction in which the NULL point is applied can be determined by generating a random number. The equation according to the equation (8) can be usually set by the number of the microphones 20a. Therefore, the number of directions to which the NULL point is applied includes a term (one) for controlling the transfer function in the target direction from the number of microphones 20a and a term (8) for controlling the difference in the transfer function from the target direction. The number can be subtracted. However, the solution of equation (8) transfer characteristics of the filter to be obtained is the number of equations with G _n be smaller than the number of _(omega), the transfer characteristic of the filter by calculating the minimum norm solution G _{n (omega)} Can be requested.

Specifically, the equation (8) can be expressed as the following equation (9). Here, the direction d1~d8 other than the target direction 2-9, adds the transfer function difference h _d is replaced each expression specific transfer function.

By determining the filter vector G ₁ ~G _N by solving the simultaneous equations shown in equation (9), it is possible to create a filter group 10.

  Here, the direction in which the NULL point is applied can be selected such that the maximum side lobe level becomes a predetermined value by monitoring the side lobe of the beam by calculation using the transfer function displayed on the contour map.

  FIG. 5 is a diagram illustrating a flowchart of a method for creating a filter for a beam forming apparatus according to the present embodiment. This flowchart can be implemented by, for example, a computer and a computer program. However, some or all of them may be implemented by an electronic circuit.

  Hereinafter, the manufacturing method of the beam forming apparatus according to the present embodiment will be described with reference to the flowchart of FIG.

  The initial value of the maximum side lobe level SLL is input (S500). The initial value of the maximum side lobe level SLL can be set to -10 dB with respect to the gain (0 dB) of the main lobe, for example. The initial value of the maximum side lobe level SLL is input by inputting it to an input device of a computer. The initial value of the maximum side lobe level SLL is input here, and the set value of the maximum side lobe level SLL is relaxed as necessary, as will be described later.

  The maximum side lobe level SLL is set (S501). The maximum side lobe level SLL is set by reflecting the initial value of the maximum side lobe level SLL input in step S500 in the computer program.

Inputting the initial value of the addition the transfer function difference _{h d} (S502). For each non-target direction d1~d8 object direction near to enter the initial value of the addition the transfer function difference h _d from the input device of the computer. To enter here is the initial value of the addition the transfer function difference h _d, the set value of the sum transfer function difference h _d, as described later, to relax as needed.

Setting the sum transfer function difference _{h d} (S503). Setting summing transfer function difference h _d is performed by reflecting the initial value of the addition the transfer function difference h _d input in step S502 to the computer program.

  A transfer function relating to all virtual sound source positions in the target range is calculated, NULL points are generated with random numbers and applied to the generation of the filter group 10, and the filter group 10 is generated according to Expression (8) (S504, transfer function calculation stage) , Filter group creation stage). Here, of the non-target directions, the direction to which the NULL point is applied is generated by a random number, applied to an equation (formula (8)) for generating the filter group 10, and the filter group 10 is generated by calculation. Each filter group 10 is generated with the direction as the target direction for all virtual sound sources in the target range.

  A beam is calculated using the filter group 10 generated in step S504, and the maximum sidelobe level SLL of the calculated beam is extracted (S505, beam calculation stage). Here, the calculation of the beam is performed as follows.

  A transfer function H (Ω) is obtained by calculation, and among the filter vectors g related to the filter group 10 generated in step S504, the filter group 10 having the target direction (θ, φ) as (0, 0) is used. The directivity vector d is calculated according to 3). For the directional characteristics D in each direction (the direction of M virtual sound sources) constituting the directional characteristic vector d, the absolute value of the sum of the directional characteristics D is calculated. Thereby, a beam which is a set of absolute values of the directivity characteristics D in each direction can be calculated.

  During actual beam forming, a sound wave signal from the sound source 1 to be detected is detected by each microphone 20a, and each of the plurality of filter groups 10 having different target directions is subjected to arithmetic processing, whereby a set of gains in each direction. To obtain a beam consisting of However, the beam obtained by the calculation obtained in step S505 is processed by one filter group 10 in which the transfer functions H (Ω) from the plurality of virtual sound sources to each microphone 20a are set to the target direction (0, 0). This is replaced by the calculated beam.

  Note that the filter group 10 used to obtain a calculated beam may be the filter group 10 having a direction other than (0, 0) as a target direction.

  It is determined whether the side lobe level SLL extracted in step S505 is equal to or lower than the maximum side lobe level SLL set in step S501 (S506, filter adjustment stage). When the extracted side lobe level SLL is equal to or lower than the set maximum side lobe level SLL, the generation of the filter group 10 is completed, and the flowchart ends. On the other hand, if the extracted side lobe level SLL is greater than the set maximum side lobe level SLL, the process proceeds to step S507.

  It is determined whether or not the number of loops is less than or equal to a preset maximum number of loops (S507, filter adjustment stage). If it is determined that the number of loops is less than or equal to the set maximum number of loops, a NULL point newly generated with a random number is determined. This is applied to generation of the filter group 10 (S504, filter adjustment stage). Then, the calculated maximum side lobe level SLL of the beam is extracted (S505, filter adjustment stage), and it is determined whether or not the extracted side lobe level SLL is equal to or lower than the set maximum side lobe level SLL (S506, filter adjustment stage). The maximum number of loops can be 2000, for example.

  As a result of performing the loop of step S505 to step S507, when the extracted side lobe level SLL is equal to or lower than the set maximum side lobe level SLL (S506, YES), the generation of the filter group 10 is completed, and the flowchart ends. To do.

If sidelobe level SLL extracted even exceeds the maximum number of loops set the does not become less than the maximum side lobe level SLL set in (S506, NO), relaxed to a smaller value the value of the addition the transfer function difference h _d set (S508, filter adjustment stage).

Adding transfer function difference h _d applied to the generation of the filter group 10 determines whether the minimum sum transfer function difference h _dmin is greater than the preset (S509, filter adjustment step), adding the transfer function difference h _d is the minimum sum transfer If it is larger than the function difference _hdmin (S509, YES), the addition transfer function difference _hd is reset to the value updated in step S508 (S503, filter adjustment stage). The reset addition transfer function difference _hd and the newly generated NULL point are applied to the generation of the filter group 10 (S504), and the maximum sidelobe level SLL of the calculated beam is re-extracted (S505). When the re-extracted side lobe level SLL is equal to or lower than the set maximum side lobe level SLL (S506, YES), the generation of the filter 10a is completed, and the flowchart ends.

On the other hand, if the addition transfer function difference _{h d} is less than or equal to the minimum sum transfer function difference _{h dmin} (S509, NO), updated by mitigating the value of the maximum sidelobe level SLL set to a large value (S510, filter adjustment step ). Then, step S501 (filter adjustment stage), step S502 (filter adjustment stage), and steps S503 to S505 are performed, and the updated maximum sidelobe level SLL is compared with the sidelobe level SLL re-extracted in step S505 ( S506). When the re-extracted side lobe level SLL is equal to or lower than the set maximum side lobe level SLL (S506, YES), the generation of the filter group 10 is completed, and the flowchart ends. When the re-extracted side lobe level SLL is equal to or lower than the set maximum side lobe level SLL (S506, NO), the loop of step S507 to step S507 is repeatedly performed.

According to the beam forming apparatus manufacturing method and the beam forming apparatus according to the present embodiment, in addition to the effects of the first embodiment, the following effects can be obtained.
Adjust the filter group so that the calculated maximum sidelobe level of the beam is reduced. Accordingly, since the beam width and the side lobe of the beam can be optimized only by calculation without using an actual sound source, the detection of the sound source generation position can be realized with higher accuracy.
Further, the adjustment of the filter group is performed by at least one of changing the position where the NULL point is applied in generating the filter group and changing the magnitude of the addition transfer function difference. Thereby, since the size of the side lobe can be further reduced, the detection of the sound source generation position can be realized with higher accuracy.
Furthermore, the filter group can be adjusted efficiently by randomly changing the position where the NULL point in the non-target direction is applied.

  FIG. 6 is a diagram showing evaluation results of beamforming characteristics by a beamforming apparatus using a filter created by the method for creating a filter for a beamforming apparatus according to the second embodiment of the present invention and a conventional beamforming apparatus. . FIG. 6A shows a contour map of beam forming characteristics by the conventional beam forming method, and FIG. 6B shows a contour map of beam forming characteristics by the beam forming method according to the present embodiment.

  The evaluation of beam forming characteristics was performed as follows. That is, a gain obtained by moving a sound source having a frequency of 500 Hz corresponding to a direction (θ, φ) on the contour map, performing arithmetic processing by a filter group having the same target direction (0, 0) in all directions, and adding them. Is displayed for each direction of the sound source.

  As described above, in general, beam forming displays, on a contour map, gains obtained by performing arithmetic processing on a sound wave signal from a sound source to be detected by a plurality of filter groups having different target directions as beams. . However, the evaluation of the beam main lobe width and side lobe can be replaced by the evaluation method used in this embodiment.

  The calculation by the filter group was performed as follows in order to compare the conventional beam forming apparatus and the beam forming apparatus using the filter created by the method for creating the filter for the beam forming apparatus according to the present embodiment under the same conditions. . In other words, in the present embodiment, a general solution of the filter is obtained from the first to ninth rows of the matrix of the formula (7), and the calculation method in the conventional beam forming apparatus is solved using the indefinite terms for the tenth to thirty-second rows. did. In addition, the censoring conditions for singular value decomposition were also the same.

  As shown in FIG. 6, according to the manufacturing method of the beam forming apparatus according to the present embodiment, the width of the main lobe of the beam on the contour map is narrowed even if the sound wave to be detected is low frequency (500 Hz). It has been proved that a peak can always be provided in the target direction. Further, since the side lobe is suppressed, it has been demonstrated that the width of the main lobe of the beam can be narrowed and the size of the side lobe can be optimized to be small.

10 filter groups,
10a filter,
20 microphone array,
20a microphone,
300 Transfer function calculator (transfer function calculator),
301 filter group creation unit (filter group creation means, filter adjustment means),
302 Beam calculation unit (beam calculation means),
303 Random number generator (filter adjustment means).

Claims (8)

A filter that emphasizes the detection signal of the target direction estimated as the direction of the sound source by adding together the detection signals calculated by a plurality of filters to which detection signals from a plurality of microphones installed in space are respectively input. A method of creating a filter that has a plurality of groups and obtains a directivity characteristic of a sound wave by calculating each of the detection signals by the filter groups having different target directions.
A transfer function calculating step of calculating each transfer function from the plurality of virtual sound source positions to the microphone based on a distance from the plurality of virtual sound source positions where the sound source may exist and the frequency of the sound source;
From the first transfer function sum obtained by calculating and adding each other in the filter so that the transfer functions from the virtual sound source position in the target direction to the microphone out of the plurality of virtual sound source positions are in phase with each other, A second transfer function sum obtained by calculating each transfer function from the virtual sound source position in the direction other than the target direction to the microphone among the plurality of virtual sound source positions and adding them to each other or the second transfer function A filter group creating stage for creating the filter group in which the filter is configured to satisfy a first condition in which a value obtained by subtracting one of the absolute values of the transfer function sum is greater than 0;
A method for producing a filter for a beam forming apparatus, comprising: A beam formed by aggregating absolute values of transfer function sums obtained by calculating each of the transfer functions from at least one of the virtual sound source positions to the microphone with the filter group. A beam calculation stage for calculating
A filter adjustment step of adjusting the filter so that the maximum value of the side lobe level of the beam calculated in the beam calculation step is lowered;
The method for producing a filter for a beam forming apparatus according to claim 1, further comprising: The filter group creating step creates the filter group so as to further satisfy a second condition in which the second transfer function sum for the virtual sound source in any one or more directions other than the target direction is 0,
In the filter adjustment step, the virtual sound source position other than the target direction with the second transfer function sum in the second condition being 0 is changed, or the second transfer function is changed from the first transfer function sum in the first condition. The filter for a beamforming apparatus according to claim 2, wherein the filter is adjusted by at least one of reducing a value obtained by subtracting either a sum or an absolute value of the second transfer function sum. How to make.   The filter for a beam forming apparatus according to claim 3, wherein the filter adjustment step randomly changes a virtual sound source position other than the target direction in which the second transfer function sum in the second condition is zero. How to make. A filter that emphasizes the detection signal of the target direction estimated as the direction of the sound source by adding together the detection signals calculated by a plurality of filters to which detection signals from a plurality of microphones installed in space are respectively input. An apparatus for creating a filter for an apparatus, which has a plurality of groups, and obtains a directivity characteristic of sound waves by calculating each of the detection signals by the filter groups having different target directions,
Transfer function calculating means for calculating each transfer function from the plurality of virtual sound source positions to the microphone based on the distance from the plurality of virtual sound source positions where the sound source can exist to the microphone and the frequency of the sound source;
From the first transfer function sum obtained by calculating and adding each other in the filter so that the transfer functions from the virtual sound source position in the target direction to the microphone out of the plurality of virtual sound source positions are in phase with each other, A second transfer function sum obtained by calculating each transfer function from the virtual sound source position in the direction other than the target direction to the microphone among the plurality of virtual sound source positions and adding them to each other or the second transfer function A filter group creating means for creating the filter group in which the filter is configured to satisfy a first condition that a value obtained by subtracting one of the absolute values of the transfer function sum is larger than 0;
An apparatus for creating a filter for a beam forming apparatus. A beam formed by aggregating absolute values of transfer function sums obtained by calculating each of the transfer functions from at least one of the virtual sound source positions to the microphone with the filter group. A beam calculating means for calculating
Filter adjusting means for adjusting the filter so that the maximum value of the side lobe level of the beam calculated in the beam calculating step is lowered;
The filter forming apparatus for a beam forming apparatus according to claim 5, further comprising: The filter group creating means creates the filter group so as to further satisfy a second condition that the second transfer function sum for the virtual sound source in any one or more directions other than the target direction is 0,
The filter adjusting means changes a virtual sound source position other than the target direction in which the second transfer function sum in the second condition is 0, or changes the second transfer function from the first transfer function sum in the first condition. The filter for a beamforming apparatus according to claim 6, wherein the filter is adjusted by at least one of reducing a value obtained by subtracting either a sum or an absolute value of the second transfer function sum. Creation device.   The filter for a beam forming apparatus according to claim 7, wherein the filter adjustment unit randomly changes a virtual sound source position other than the target direction in which the second transfer function sum in the second condition is zero. Creation device.