JP2019050492A - Filter coefficient determining device, filter coefficient determining method, program, and acoustic system - Google Patents

Abstract

To reliably determine a filter coefficient that is more optimum.SOLUTION: In order to determine a filter coefficient set to a digital filter arranged on a stage preceding a speaker element of a cylindrical array, a directivity basis is calculated, which serves as a basis of the directivity in sound outputted from the cylindrical array. Then, a calculation is performed with the use of an expansion coefficient obtained by applying desired directivity to a filter basis, which is a function used for calculating the directivity basis, as well as to the directivity basis. The present technique is applicable to, for example, an acoustic system equipped with the cylindrical array.SELECTED DRAWING: Figure 5

Description

  The present disclosure relates to a filter coefficient determination device, a filter coefficient determination method, a program, and an acoustic system, and in particular, a filter coefficient determination device, a filter coefficient determination method, a program that enables more optimal filter coefficients to be reliably determined. And acoustic systems.

  In recent years, voice control terminals used at home have been provided by various companies. These voice control terminals are equipped with a microphone array for recognizing voice, and can recognize, for example, speech in a noisy environment or speech from a distance from the speech control terminal.

  In addition, when outputting a voice, music, or the like, for example, the voice control terminal is configured to be reproduced toward all directions around 360 degrees around it. Furthermore, a voice control terminal has also been developed, which is configured to make it possible to hear a wide range or only in a specific range in a predetermined high-pitched sound.

  In general, a speaker that emits sound strongly only in a certain direction is called a directional speaker, and conventionally, directional speakers of various configurations have been considered. Moreover, as a principle which forms directivity in a directional speaker, it can be divided roughly into two, for example. The first principle is used in a parametric array that uses ultrasound to form directivity, and the second principle is that a plurality of ordinary speakers are arranged to separate the wave amplitude and phase in space for each frequency. It is used by the speaker array which forms directivity by controlling to.

  Here, in Non-Patent Document 1, the inventors of the present application disclose a technique for designing a filter that enables three-dimensional directivity control using a circular speaker array. Further, as disclosed in Non-Patent Document 2, as a method of calculating filter coefficients, a method of performing spatial Fourier transform once is known. Further, Non-Patent Document 3 discloses a method of superimposing multipole sound sources based on spherical harmonics expansion in order to synthesize arbitrary directional characteristics using a speaker array.

Koji Sato, Yoichi Haneda "Three-dimensional directivity control based on circular harmonic expansion mode using circular array," Technical Report of the Faculty of Engineering, vol. 116, no. 475, EA 2016-119, pp. 213-218, 2017 March. Haneda Yoichi "Wavenumber and Array Signal Processing", Proceedings of the 2017 Spring Meeting of the Acoustical Society of Japan, 0122_1-8-15haneda 2017 spring SS. Pdf, March 2017. Haneda Yoichi, Furuya Kenichi, Shimauchi Suetsu "Directivity synthesis using multipole sound source based on spherical harmonics expansion", Journal of the Acoustical Society of Japan, 69, 11, pp. 577-588, November 2013.

  By the way, in the speaker array as described above, studies are made on how to arrange the speakers and how to design the filter coefficients. Conventionally, the arrangement of the speakers can be divided into three, a linear array in which the speakers are linearly arranged, a circular array in which the speakers are circularly arranged, and a spherical array in which the speakers are spherically arranged.

  For example, since the linear array is indistinguishable in the vertical direction, in controlling its directivity, a method of fixing directivity in the axial direction of a straight line called an endfire array is adopted. It was difficult to change the direction arbitrarily. In addition, while circular arrays can freely rotate directivity in the horizontal plane, it has been very difficult to distinguish in the vertical direction. In addition, although the spherical array can reproduce directivity in the vertical direction and the left and right direction, it may be difficult to install because it is a sphere.

  Therefore, for example, in a speaker array that can be easily installed, it is necessary to reliably determine a more optimal filter coefficient in order to be able to reproduce directivity in the vertical and horizontal directions. Become.

  The present disclosure has been made in view of such a situation, and is intended to be able to reliably determine a more optimal filter coefficient.

  A filter coefficient determination device according to one aspect of the present disclosure is a cylinder in which a circular array in which two or more speaker elements are circumferentially arranged is stacked in two or more stages in the vertical direction orthogonal to the circumferential direction. Filter coefficient determining device for determining a filter coefficient to be set in a digital filter installed in a front stage of the speaker element constituting an antenna array, which is a directivity based on directivity of sound output from the cylindrical array Directivity basis calculating unit for calculating a sexual basis, a filter basis that is a function used to calculate the directional basis, and an expansion obtained by giving desired directivity to the directional basis And a filter coefficient calculator for calculating the filter coefficients of the cylindrical array using the coefficients.

  A filter coefficient determination method according to one aspect of the present disclosure is a cylinder in which a circular array in which two or more speaker elements are circumferentially arranged is stacked in two or more stages in the vertical direction orthogonal to the circumferential direction. A filter coefficient determination device for determining filter coefficients to be set in a digital filter installed in front of the speaker elements constituting the matrix array is a directional basis on which directivity of sound output from the cylindrical array is based Using the filter base which is a function used to calculate the directional base, and the expansion coefficient obtained by giving the desired directivity to the directional base, Computing the filter coefficients of the cylindrical array.

  A program according to one aspect of the present disclosure is a cylindrical array in which a circular array in which two or more speaker elements are arranged in a circumferential direction is stacked in two or more stages in a vertical direction orthogonal to the circumferential direction. The directivity factor as the basis of the directivity of the sound output from the cylindrical array is provided to the computer of the filter coefficient determination device which determines the filter coefficient to be set in the digital filter installed in the front stage of the speaker element to be configured. The cylinder using the calculation, a filter basis which is a function used to calculate the directional basis, and an expansion coefficient obtained by giving a desired directivity to the directional basis. Computing the filter coefficients of the form array.

  An acoustic system according to one aspect of the present disclosure is a cylindrical array in which a circular array in which two or more speaker elements are arranged in a circumferential direction is stacked in two or more stages in a vertical direction orthogonal to the circumferential direction. And a digital filter placed in front of the speaker element constituting the cylindrical array, and calculating a directivity base which is a basis of the directivity of the sound output from the cylindrical array, the directivity Calculate the filter coefficients of the cylindrical array using the filter basis which is a function used to calculate the basis and the expansion coefficient obtained by giving the desired directivity to the directional basis The filter coefficients determined by the above are set in the digital filter.

  In one aspect of the present disclosure, a directional basis on which directivity of sound output from a cylindrical array is based is calculated, and a filter basis that is a function used to calculate the directional basis; The filter coefficients of the cylindrical array are determined by calculation using the expansion coefficients determined by giving the desired directivity to the sex basis.

  According to an aspect of the present disclosure, more optimal filter coefficients can be reliably determined.

  In addition, the effect described here is not necessarily limited, and may be any effect described in the present disclosure.

It is a block diagram showing an example of composition of an embodiment of an acoustic system using filter coefficients determined by applying this art. It is a figure which shows one structural example of a cylindrical array. It is a figure explaining the control point in a polar coordinate system. It is a conceptual diagram of a multipole sound source. It is a block diagram which shows the structural example of a filter factor determination apparatus. It is a flow chart explaining filter coefficient determination processing. Fig. 21 is a block diagram illustrating a configuration example of an embodiment of a computer to which the present technology is applied.

  Hereinafter, specific embodiments to which the present technology is applied will be described in detail with reference to the drawings.

<Example of configuration of sound system>
FIG. 1 is a block diagram showing a configuration example of an embodiment of an acoustic system using filter coefficients determined by applying the present technology.

As shown in FIG. 1, the acoustic system 11 includes a cylindrical array 21 having a reproduction device 12, L filters 13-1 to 13-L, and L speaker elements 14-1 to 14-L. Is configured. Further, when determining the filter coefficient, assuming that a spherical surface larger than the size of the cylindrical array 21 is output from the speaker elements 14-1 to 14-L on a spherical surface separated by a predetermined distance or more from the cylindrical array 21. M control points Ω _{1 to} Ω _{M, which} are microphones or the like for observing the sound to be

  The reproduction device 12 reproduces predetermined sound source data, and an acoustic signal for outputting the reproduced sound from the speaker elements 14-1 to 14-L is a speaker element via the filters 13-1 to 13-L. 14-1 to 14-L.

The filters 13-1 to 13-L are disposed in front of the speaker elements 14-1 to 14-L, respectively, and filter coefficients w ₁ (ω) to w _L (ω) set for each of them are The output sound signal is multiplied and output.

  The speaker elements 14-1 to 14-L are audible range speakers capable of outputting sound in a frequency band that human beings can hear as sound, and the acoustic signal supplied from the reproduction device 12 through the filters 13-1 to 13-L. Output a sound according to And in the acoustic system 11 of this embodiment, the cylindrical array 21 as shown in FIG. 2 is configured by the speaker elements 14-1 to 14-L.

In the cylindrical array 21 shown in FIG. 2, the plurality of speaker elements 14 are represented by circles. In the cylindrical array 21, a circular array in which a predetermined number of speaker elements 14 are arranged at equal intervals in the circumferential direction is two steps in the vertical direction (Z direction in FIG. 2) orthogonal to the circumferential direction. It is laminated and comprised. In the example of FIG. 2, the cylindrical array 21 is configured by stacking three or more stages of circular arrays in which eight speaker elements 14 are arranged at 45 degree intervals in the circumferential direction. In the following description, the cylindrical array 21 has a finite number of stages (h _z = 1, 1) in the positive direction of the Z axis with the circular array disposed at the center in the Z axis direction as a reference (h _z = 0). And the circular array of finite stages (h _z = −1,...) Arranged in the negative direction of the Z axis.

  The cylindrical array 21 having such a configuration is, for example, easy to install as compared with a spherical array, and can direct directivity not only in the horizontal plane but also in the vertical direction.

Then, the acoustic system 11 of FIG. 1 sets the filter coefficients w ₁ (ω) to w _L (ω) in the filters 13-1 to 13 -L appropriately, so that the cylindrical array 21 only in a predetermined direction. It is possible to provide directivity from which the sound strongly radiates.

<Method of Determining Filter Coefficient in Conventional Cylindrical Array>
First, the filter design of the cylindrical array 21 will be described in the prior art.

  Conventionally, as shown in FIG. 1, a plurality of control points Ω are arranged at positions where sound is observed to control sound pressure, and filter coefficients are designed so as to match the directivity for which the sound pressure is targeted. Methods are used.

For example, in a speaker array having L speaker elements 14-1 to 14-L, a matrix having as an element the transfer function g _ps (ω) from the s-th speaker element 14-s to the p-th control point Ω _p The transfer function matrix G (ω) which is is expressed as shown in the following equation (1). Also, a desired sound pressure obtained by vectorizing the filter coefficient vector w (ω) of the angular frequency ω and the characteristics d ₁ (ω) to d _M (ω) of the desired sound pressure at _M control points Ω _{1 to} Ω _M The vector d (ω) is also expressed as shown in the following equation (1).

  At this time, the filter coefficient vector w (ω) can be obtained by computing the following equation (2).

  However, as shown in the equation (2), since the calculation involves the calculation of the inverse matrix, the filter coefficient vector w (ω) is inevitably too large, and as a result, the output of the filter 13 becomes unstable. It becomes easy to become.

  Therefore, in general, it is possible to prevent the filter coefficient vector w (ω) from becoming too large by performing the operation shown in the following equation (3).

  Here, in equation (3), β is a small positive number, and I is a square matrix. The value of β is generally obtained by trial and error, and for example, a large value is required at low frequencies.

On the other hand, as disclosed in Non-Patent Document 2 described above, a method of performing spatial Fourier transform once is known as a method of calculating the filter coefficients w ₁ (ω) to w _L (ω). In this method, for example, for a circular array, it is possible to specify which order increases the gain (magnitude) of the filter 13 by expanding with circular harmonics. In this way, it is possible to prevent the entire output of the filter 13 from becoming unstable by previously determining which order to use for the order of the circular harmonic function for each frequency.

However, while this method can only be used when the transfer function g _ps (ω) is analytically specified, it should be used when the transfer function g _ps (ω) is not specified. I could not

Therefore, in the filter design of the cylindrical array 21 in the present embodiment, first, taking the filter design with a spherical array whose transfer function g _ps (ω) is theoretically known as an example, the filters for each conventional order are used A method of obtaining the coefficients w ₁ (ω) to w _L (ω) is adopted. This method is generally called mode matching.

For example, as shown in FIG. 3, the characteristic d (Ω _p ) of the desired sound pressure at the control point Ω _p = (θ _p , φ _p ) on the spherical surface with radius r has an expansion coefficient d _nm and spherical harmonics Y _It is calculated | required by following Formula (4) using ^nm ((ohm) _p ). Here, θ and φ represent angles in the polar coordinate system, and Ω represents the overall coordinate position.

Here, for the M control points Ω _{1 to} Ω _M , the characteristic d (Ω _p ) of the desired sound pressure is expanded by the spherical harmonic function Y _n ^m (Ω _p ) (p = 1, 2,..., M ). Furthermore, for the L filter coefficients w ₁ (ω) to w _L (ω), the speaker position Ω _S = (θ _S , φ _S ) of the radius a of the spherical array is placed in front of the speaker element 14 expand the filter coefficients of the filter 13 w (Ω _S) at spherical harmonics ^{_{Y n m (Ω S) (}} s = 1,2, ···, L) is. Thus, the filter coefficient w (Ω _s ) is obtained by the following equation (5).

At this time, the transfer function G (Ω _p | Ω _S ) from the speaker position Ω _S = (θ _S , φ _S ) to the control point Ω _p = (θ _p , φ _p ) of the spherical array is the spherical harmonic function Y _n It is expressed as shown in the following equation (6) using ^m (Ω _p ).

Here, when the sound pressure observed when the filter coefficient w (Ω _s ) shown in the equation (5) is given to the speaker array is controlled to match the characteristic d (Ω _p ) of the desired sound pressure, The following equation (7) holds.

  If the relational expression shown by such Formula (7) and the above-mentioned Formula (4) are compared by the same order about the order n and the order m of a circular harmonic function, it will become like following Formula (8).

Here, the expansion coefficient d _nm is obtained from the above equation (4), and the coefficient b _nm is theoretically derived, so the filter coefficient w (Ω _s ) computes the above equation (5) It can be determined by

  <Method of Determining Filter Coefficients in Cylindrical Array of the Present Technology>

By the way, in the cylindrical array 21 used in the present embodiment, it is fundamental to design filter coefficients in the space Fourier transform domain. In general, when performing filter design in the space Fourier transform domain, it is necessary to theoretically specify the transfer function G (Ω _p | Ω _S ) from the speaker element 14 to the control point Ω. For example, theoretically considering a cylindrical array, it is usually required to be infinitely connected in the height direction (z direction). However, in an actual cylindrical array, as described with reference to FIG. 2, the theoretical value can not be used because the height direction is finite.

Therefore, in the following, for example, as shown in FIG. 1, the control points Ω _{1 to} Ω _M are arranged, and the sound is output from the speaker elements 14-1 to 14-L, whereby the transfer function G (Ω _p | Ω _S ) And calculate the basis necessary to obtain the filter coefficient from its transfer function G (Ω _p | Ω _S ), and use this to determine the filter coefficient for each order n and m of circular harmonics The method will be described. That is, even if the transfer function G (Ω _p | Ω _S ) can not be analytically expressed and theoretically the coefficient b _nm can not be derived, the optimal filter coefficient is determined with certainty. be able to.

For example, if the desired sound pressure characteristic d (Ω _p ) is developed using a directional basis _An ^m (Ω _p ) as shown in the following equation (9) in the conventional method as described above, the desired sound is obtained The pressure characteristic d (Ω _p ) is expressed as shown in the following equation (10).

Comparing the equation (10) and the last equation of the equation (7), the expansion coefficient a _nm can be regarded as identical to the expansion filter coefficient w _nm (a _nm = w _nm ) , And expansion filter coefficients w _nm can be obtained directly.

Therefore, in the present embodiment, the directivity calculated by measuring transfer function G (Ω _p | Ω _S ) in advance even if directional base _An ^m (Ω _p ) is unknown. with basal _a ^{n m} (Ω _p), proposes a method for determining the filter coefficients w (Ω _S) used in the cylindrical array 21. In the following, the coordinate position of the speaker position Ω _S is represented not by the polar coordinate system (θ _S , φ _S ) but by the cylindrical coordinate system (φ _S , z _S ).

First, on the premise that the directional basis _An ^m (Ω _p ) is found, the filter basis E _n ^m (Ω _S ) that develops the filter coefficients is assumed to be as shown by the following equation (11) Based on this, it is assumed that the function can be described analytically as described later.

Further, the characteristic d (Ω _p ) of the desired sound pressure can be expressed by the following equation (12) with reference to the method using the conventional spherical harmonic function Y _n ^m (Ω _p ) as described above . In the following, in consideration of extensibility, the range of the order n is not n> 0, and the range of the order n and the order m is expanded from −∞ to ∞. Similarly, the filter coefficient w (Ω _s ) can be expressed by the following equation (13), and the transfer function G (Ω _p | Ω _s ) can be expressed by the following equation (14).

Here, since the reverse expansion of the equation (14) described above is the following equation (15), if the transfer function G (Ω _p | Ω _S ) can be observed based on this equation (15), The directional basis _An ^{m m} (Ω _p ) can be determined.

Therefore, first, it is necessary to define the filter basis E _n ^m (Ω _S ) in the cylindrical array 21. For example, the filter basis E _n ^m (Ω _s ) can be defined using either of the two functions described below.

First, there is a method of defining a filter basis E _n ^m (Ω _S ) using a plane wave basis which has the same form as a general solution of a wave equation in a cylindrical coordinate system. For example, when the sound source is a continuum, the filter basis E _n ^m (Ω _S ) with reference to the general solution in the cylindrical coordinate system has an angle φ _{S in the} circumferential direction, an order (wave number) m in the φ _S direction, and z The axial wavenumber k _z and the height z _S on the z axis are expressed by the following equation (16).

Usually, in the cylindrical array 21, the speaker elements 14 are discretely arranged. For example, as shown in FIG. 2 described above, H _phi pieces in a direction (circumferential direction) phi, and, H _z stage speaker device 14 is disposed a cylindrical array 21 configured by Δz intervals in the z-axis direction Filter base E _n ^m (Ω _S ) is expressed by the following equation (17).

Further, in the method of defining the filter basis E _n ^m (Ω _s ) using multiple poles, the filter basis E _n ^m (Ω _s ) is expressed by the following equation (18).

Therefore, in the present embodiment, directivity is obtained by selecting one of the equations (17) and (18), combining the order n and the order m, and determining the sound to be given to the respective speaker elements 14. base _a ^{n m} _{(Ω p)} can be calculated based on equation (15) described above.

Here, as shown in the above-mentioned equation (17), the circumferential angle .phi. _H in the polar coordinate system (see FIG. 3) and the Z-axis direction of the circular array (see FIG. 2) constituting the cylindrical array 21. The relationship with the column h _z will be described.

For example, a circular array H _phi number of speakers element 14 in the horizontal direction is arranged, are arranged in a number of stages H _z in the Z-axis direction, the sum, the cylindrical array 21 consisting of _H φ × H _z number of speakers element 14 The position of each speaker element 14 can be defined as (a, φ _hφ , z _hz ) in a cylindrical coordinate system of radius a.

At this time, φ _hφ is expressed by the following equation (19), and z _hz is expressed by the following equation (20) using the speaker spacing Δz in the z-axis direction.

In addition, _hφ used in equation (19) is as shown in the following equation (21), and h _z used in equation (20) is the following equation (22) using a floor function: As shown in).

In addition, as shown in the above-mentioned equation (18), a method of defining the filter basis E _n ^m (Ω _S ) using multiple poles will be described. Here, as shown in FIG. 2, a description will be given using a cylindrical array 21 configured by stacking three stages of circular arrays in which the speaker elements 14 are arranged at equal intervals in the circumferential direction in the Z direction. .

For example, assuming that a circular array arranged at each stage of the cylindrical array 21 is one point sound source, and a spherical harmonic function Y _n ^m (θ, φ) is expressed by a combination of multipole sound sources, as shown in FIG. It can be described using a conceptual diagram of a multipole sound source as shown in FIG. In FIG. 4, the monopole sound source is spatially differentiated, and in the vertical quadrupole, the sound source is expressed in common.

First, in order to express the spherical harmonic _Y ^{0 0} (theta, phi) is spherical harmonics _Y ^{0 0} (theta, phi) is more to be a following expression (23), indicated by A in FIG. 4 It can be seen that it is sufficient to drive the point sound source arranged at the origin with √1⁄4π.

Similarly, in consideration of the speaker interval (2Δz) as shown in B of FIG. 4, a driving coefficient for expressing the spherical harmonic function Y ₁ ⁰ (θ, φ) can be obtained.

Further, since the spherical harmonic function Y ₂ ⁰ (θ, φ) is expressed by the following equation (24), it is expressed by the combination of the vertical quadrapole shown in C of FIG. 4 and the monopole sound source of A of FIG. can do.

Summarizing the above, when there are three point sound sources on the z-axis, spherical harmonics Y _n ⁰ (θ, φ) up to order n = 2 can be represented, and their respective drive coefficients F _n ( h _z , ω) are as shown in the following equations (25) to (27).

Here, C ₁ used in Formula (26) is as shown in the following Formula (28), and C ₂ used in Formula (27) is as shown in the following Formula (29). Also, the elements of the driving coefficient correspond to h _z = 1, 0, -1 from the left.

  Note that how to obtain these driving coefficients is described in detail in the above-mentioned Non-Patent Document 3.

Then, if the directional basis _An ^m (Ω _p ) can be determined, the desired directivity (characteristic d (Ω _p ) of desired sound pressure) is given, and the expansion is performed based on the above equation (12). The coefficient a _nm can be identified. Thus, the filter coefficient w (Ω _s ) is the filter basis E _n ^m (Ω _s ) used to determine the expansion coefficient a _nm specified in this way and the directional basis A _n ^m (Ω _p ) It is calculated | required by calculating Formula (13) mentioned above using.

<Configuration Example of Filter Coefficient Determination Device>
FIG. 5 is a block diagram showing a configuration example of a filter coefficient determination device to which the above-described filter coefficient determination method is applied.

  As shown in FIG. 5, the filter coefficient determination device 31 includes a filter basis selection unit 32, a directivity basis calculation unit 33, an expansion coefficient specification unit 34, a filter coefficient calculation unit 35, and a filter coefficient setting unit 36. Ru.

The filter basis selection unit 32 selects the filter basis E _n ^m (Ω _S ) used to determine the filter coefficient w (Ω _S ) in the filter coefficient determination device 31. For example, the filter basis selection unit 32 may use a filter basis E _n ^m (Ω _S ) based on plane wave expansion as shown in the above equation (17), and multipole expansion as shown in the above equation (18) One of the filter bases E _n ^m (Ω _S ) based on is selected. Then, the filter basis selection unit 32 supplies the selected filter basis E _n ^m (Ω _S ) to the directivity basis calculation unit 33 and the filter coefficient calculation unit 35.

The directivity basis calculation unit 33 uses the transfer function G (Ω _p | Ω _S ) obtained by measurement and the filter basis E _n ^m (Ω _S ) supplied from the filter basis selection unit 32 as described above. The directional basis _An ^m (Ω _p ) is calculated by computing equation (15). Then, the directivity basis calculation unit 33 supplies the calculated directivity basis _An ^m (Ω _p ) to the expansion coefficient specification unit 34.

The expansion coefficient specification unit 34 gives desired directivity (characteristic d (Ω _p ) of desired sound pressure) to the directional basis _An ^m (Ω _p ) supplied from the directional basis calculation unit 33. The expansion coefficient a _nm can be specified based on the above equation (12). Then, the expansion coefficient identification unit 34 supplies the identified expansion coefficient a _nm to the filter coefficient calculation unit 35.

The filter coefficient calculation unit 35 uses the filter basis E _n ^m (Ω _S ) supplied from the filter basis selection unit 32 and the expansion coefficient a _nm supplied from the expansion coefficient identification unit 34 to obtain the above equation (13). Calculate). Thereby, the filter coefficient computing unit 35 obtains the filter coefficient w (Ω _S ) and supplies the filter coefficient w (Ω _S ) to the filter coefficient setting unit 36.

At this time, the filter coefficient calculation unit 35 calculates the filter coefficient w (Ω _s ) on the basis of circular harmonic development for filter coefficient expansion in the circumferential direction. Then, for example, when the filter basis E _n ^m (Ω _s ) of Expression (17) is selected by the filter basis selection unit 32, the filter coefficient computation unit 35 performs plane wave development for filter coefficient development in the vertical direction. The filter coefficient w (Ω _s ) can be calculated on the basis of On the other hand, when the filter basis E _n ^m (Ω _s ) of equation (18) is selected by the filter basis selection unit 32, for example, the filter coefficient computation unit 35 generates multipoles for filter coefficient expansion in the vertical direction. The filter coefficient w (Ω _S ) can be calculated on the basis of the expansion.

Furthermore, when calculating the filter coefficient w (Ω _s ), the filter coefficient calculation unit 35 limits the synthesis order in the low band using L1 norm regularization as disclosed in the above-mentioned Non-Patent Document 1 By doing this, the gain of the filter 13 can be suppressed. Alternatively, L2 norm regularization can be used instead of L1 norm regularization.

Here, L1 norm regularization and L2 norm regularization will be described. First, as shown in the equation (13) described above, the filter coefficient is obtained by using the expansion coefficient a _nm obtained from the characteristic d (Ω ₀ ) of the desired sound pressure with the directional basis _An ^m (Ω ₀ ) It can be calculated. For example, since the filter coefficient base E _n ^m (Ω _S ) is a constant, the filter size (gain) depends on the expansion coefficient a _nm . L2 norm regularization is a method of limiting the sum of squares of the expansion coefficient a _nm to be a certain value λ or less, and using an absolute value of the expansion coefficient a _nm , as shown in the following equation (30) Be expressed.

On the other hand, L1 norm regularization is a method of limiting the sum of absolute values of expansion coefficients a _nm to be a certain value λ or less, and is expressed by the following equation (31).

Generally, in the case of L1 norm regularization, it is difficult to optimize, so it is not always possible to obtain an expansion solution a _nm which is an optimal solution. In that case, it is better to use L2 norm regularization.

Thus, by using either L1 norm regularization or L2 norm regularization, the expansion coefficient a _nm obtained when expanded with a directional basis _An ^m (Ω ₀ ) with poor radiation efficiency is increased. Can be prevented. As a result, the filter gain can be suppressed as a result, and natural reproduction without distortion can be performed.

The filter coefficient setting unit 36 holds the filter coefficient w (Ω _s ) supplied from the filter coefficient calculation unit 35, and according to the characteristic d (Ω _p ) of the desired sound pressure to be obtained for the cylindrical array 21. , Set an appropriate filter coefficient w (Ω _s ) in the filter 13.

  As described above, the filter coefficient determination device 31 is an optimal filter coefficient for providing the cylindrical array 21 with directivity so that sound can be output at a desired volume in the desired direction in the vertical and horizontal directions. Can be determined for sure.

Processing Example of Filter Coefficient Determination Processing
Filter coefficient determination processing by the filter coefficient determination device 31 will be described with reference to the flowchart shown in FIG.

For example, as shown in FIG. 1, by arranging control points Ω _{1 to} Ω _M and outputting sound from the speaker elements 14-1 to 14-L, the transfer function G (Ω _p | Ω _S ) is measured, and a cylinder When the desired sound pressure characteristic d (Ω _p ) to be determined for the form array 21 is set, the process is started.

In step S11, the filter basis selection unit 32 performs filter basis E _n ^m (Ω _S ) based on plane wave expansion as shown in the above-mentioned equation (17), and multiplexing as shown in the above-mentioned equation (18) One of filter bases E _n ^m (Ω _S ) based on pole expansion is selected. Then, the filter basis selection unit 32 supplies the selected filter basis E _n ^m (Ω _S ) to the directivity basis calculation unit 33 and the filter coefficient calculation unit 35.

In step S12, the directional basis calculation unit 33 calculates the filter basis E _n ^m (Ω _S ) supplied from the filter basis selection unit 32 in step S11, and the measured transfer function G (Ω _p | Ω _S ). The equation (15) described above is calculated using this. Thereby, the directional basis calculation unit 33 calculates the directional basis _An ^m (Ω _p ) and supplies the directional basis calculation unit 33 with the expansion coefficient specification unit 34.

In step S13, the expansion coefficient identifying unit 34 sets the characteristic d (Ω _p ) of the desired sound pressure set with respect to the directional basis _An ^m (Ω _p ) supplied from the directional basis calculating unit 33 in step S12. The expansion coefficient a _nm is specified based on the above equation (12) using Then, the expansion coefficient identification unit 34 supplies the identified expansion coefficient a _nm to the filter coefficient calculation unit 35.

In step S14, the filter coefficient calculator 35 receives the filter basis E _n ^m (Ω _S ) supplied from the filter basis selector 32 in step S11, and the expansion coefficient a supplied from the expansion coefficient identification unit 34 in step S13. _The above equation (13) is calculated using _nm . Thereby, the filter coefficient computing unit 35 obtains the filter coefficient w (Ω _s ) and supplies it to the filter coefficient setting unit 36, and the filter coefficient w (Ω _s ) is set in the filter 13 by the filter coefficient setting unit 36 And the process is ended.

As described above, the filter coefficient w (Ω _S ) determined by the filter coefficient determination device 31 is a linear sum of the filter bases E _n ^m (Ω _S ), as shown in the above-mentioned equation (13). Also, for example, when the combination of an order n and an order m is determined and the cylindrical array 21 is ringed using the filter bases E _n ^m (Ω _S ) one by one, the directivity observed in the periphery is It follows that the sex basis _An ^m (Ω _p ). Here, the directional basis _An ^m (Ω _p ) expresses not only the directivity pattern in each order but also the energy when emitting the sound.

  This gives the radiation efficiency when ringing the speaker element 14 with the same size. For example, if the radiation efficiency is low, it means that a loud sound must be given to each speaker element 14 to hear the sound at a certain control point. In this case, the speaker element 14 is distorted, and not only the sound deteriorates but also the directivity is not reproduced.

Therefore, in the filter coefficient determination device 31, as described above, the radiation efficiency for each of the order n and the order m can be known in advance as a directional basis _An ^m (Ω _p ). It becomes possible to make directivity without using. By making directivity in this way, the cylindrical array 21 can perform natural reproduction without distortion.

  In addition, the cylindrical array 21 has an advantage that directivity can be controlled in the vertical direction as compared with the circular array. Furthermore, the cylindrical array 21 has an advantage that it can be rotated in the φ direction while maintaining the same directivity pattern. Since this advantage can radiate the same energy in each direction when analyzing a sound field, it is useful to apply to the analysis of a sound field.

As described above, the cylindrical array 21 in the present embodiment is configured by arranging a plurality of normal speaker elements 14, and as described above, the filter coefficient w (Ω _s ) is determined to be high. It is possible to form directivity not only in the range component but also in the normal voice band. Then, by applying such a cylindrical array 21 to a voice control terminal used at home, for example, only to the user who is in conversation in the conversation with the artificial intelligence assistant service provided on the cloud, etc. It is feasible to transmit the sound as you would hear it. Also, for example, when it is recognized that the characteristic of the user's ear is deteriorated by identifying the user who is in conversation, the direction from the cylindrical array 21 toward the user is corrected so as to compensate for the characteristic. It becomes feasible to correct the frequency characteristics.

Here, for example, when plane wave expansion is performed with respect to filter coefficient expansion of the cylindrical array 21 in the vertical direction, the filter bases E _n ^m (Ω _S ) are orthogonal to each other. On the other hand, the filter bases E _n ^m (Ω _S ) in the case of multipole expansion are not orthogonal. Therefore, the relationship between the equation (14) and the equation (15) described above is not satisfied in the case of multipole expansion. However, in the present embodiment, since the transfer function G (Ω ₀ | Ω _S ) is based on measured data, this difference does not cause a major problem.

On the other hand, for the filter factor expansion of the cylindrical array 21 in the vertical direction, the directional basis A _n ^m (Ω ₀ ) in the case of using the multipole expansion is the conventional spherical harmonic function Y _n ^m (Ω ₀ ) Because of this, the coefficient b _n used in equation (6) expressing the radiation efficiency can be obtained in a pseudo manner. However, strictly, in the case of using multipole expansion, the coefficient is b _nm , and the value depends on the order n and the order m.

On the other hand, in the case of plane wave expansion, since the cylindrical coordinate system is considered, it can not be associated with the spherical harmonics Y _n ^m (Ω ₀ ). Therefore, the coefficient b _nm can not be determined. However, since the directional basis _An ^m (Ω ₀ ) itself represents the radiation efficiency, for example, calculation is performed such that the maximum value of the directional basis _An ^m (Ω ₀ ) for each order is 1 The radiation efficiency becomes clear.

Also, using the expansion coefficient a _nm obtained in the directional basis _An ^m (Ω ₀ ) in each case for the characteristic d (Ω ₀ ) of the desired sound pressure, the filter basis E _n ^m (in each case) calculated in Omega _S), the results obtained when driving the cylindrical array 21 in each resulting filter coefficients w (Ω _S) are the same. However, this is only the case where L1 norm regularization or L2 norm regularization is not considered. Therefore, the effects are the same with regard to the choice of plane wave expansion or multipole expansion.

<Example of computer configuration>
Note that the processes described with reference to the above-described flowchart do not necessarily have to be processed in chronological order according to the order described as the flowchart, and processes performed in parallel or individually (for example, parallel processes or objects Processing) is also included. The program may be processed by a single CPU or may be distributed and processed by a plurality of CPUs.

  Further, the above-described series of processes (filter coefficient determination method) can be performed by hardware or can be performed by software. When a series of processes are executed by software, the various functions are executed by installing a computer in which a program constituting the software is incorporated in dedicated hardware or various programs. The program can be installed, for example, on a general-purpose personal computer from a program recording medium on which the program is recorded.

  FIG. 7 is a block diagram showing an example of a hardware configuration of a computer that executes the series of processes described above according to a program.

  In the computer, a central processing unit (CPU) 101, a read only memory (ROM) 102, and a random access memory (RAM) 103 are mutually connected by a bus 104.

  Further, an input / output interface 105 is connected to the bus 104. The input / output interface 105 includes an input unit 106 including a keyboard, a mouse and a microphone, an output unit 107 including a display and a speaker, a storage unit 108 including a hard disk and a non-volatile memory, and a communication unit 109 including a network interface. A drive 110 for driving a removable medium 111 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is connected.

  In the computer configured as described above, for example, the CPU 101 loads the program stored in the storage unit 108 into the RAM 103 via the input / output interface 105 and the bus 104 and executes the program. Processing is performed.

  The program executed by the computer (CPU 101) is, for example, a magnetic disk (including a flexible disk), an optical disk (CD-ROM (Compact Disc-Read Only Memory), DVD (Digital Versatile Disc), etc.), a magneto-optical disk, or a semiconductor It is recorded on a removable medium 111 which is a package medium including a memory or the like, or is provided via a wired or wireless transmission medium such as a local area network, the Internet, and digital satellite broadcasting.

  The program can be installed in the storage unit 108 via the input / output interface 105 by mounting the removable media 111 in the drive 110. The program can be received by the communication unit 109 via a wired or wireless transmission medium and installed in the storage unit 108. In addition, the program can be installed in advance in the ROM 102 or the storage unit 108.

  The present embodiment is not limited to the above-described embodiment, and various modifications can be made without departing from the scope of the present disclosure. Further, the effects described in the present specification are merely examples and are not limited, and other effects may be present.

  DESCRIPTION OF SYMBOLS 11 acoustic system, 12 reproduction apparatus, 13 filter, 14 speaker element, 21 cylindrical array, 31 filter coefficient determination device, 32 filter base selection unit, 33 directivity base calculation unit, 34 expansion coefficient specification unit, 35 filter coefficient calculation unit , 36 Filter coefficient setting unit

Claims (8)

A circular array, in which two or more speaker elements are arranged in the circumferential direction, is disposed in front of the speaker elements forming a cylindrical array in which two or more stages are stacked in the vertical direction orthogonal to the circumferential direction. A filter coefficient determination device for determining a filter coefficient to be set to a digital filter to be
A directional basis calculation unit that calculates a directional basis that is a basis of the directivity of the sound output from the cylindrical array;
Filter coefficients of the cylindrical array using a filter basis, which is a function used to calculate the directional basis, and expansion coefficients obtained by giving a desired directivity to the directional basis And a filter coefficient calculation unit that calculates a filter coefficient determination unit. The filter coefficient calculation unit is based on circular harmonic development for filter coefficient expansion in the circumferential direction of the cylindrical array, and multipole expansion for filter coefficient expansion in the vertical direction of the cylindrical array. The filter coefficient determination device according to claim 1, wherein filter coefficients of the cylindrical array are calculated. The filter coefficient calculation unit is based on circular harmonic expansion for filter coefficient expansion in the circumferential direction of the cylindrical array, and planar wave expansion for filter coefficient expansion in the vertical direction of the cylindrical array. The filter coefficient determination device according to claim 1, wherein filter coefficients of the cylindrical array are calculated. The directional basis calculation unit is a transfer function obtained by observing sound output from the cylindrical array at a plurality of control points on a spherical surface separated from the cylindrical array by a predetermined distance or more, and the filter The filter coefficient determination device according to any one of claims 1 to 3, wherein the directional basis is calculated based on a filter basis on which a coefficient is based. The filter coefficient determination device according to any one of claims 1 to 4, wherein the filter coefficient calculation unit calculates the filter coefficient using L2 norm regularization or L1 norm regularization. A circular array, in which two or more speaker elements are arranged in the circumferential direction, is disposed in front of the speaker elements forming a cylindrical array in which two or more stages are stacked in the vertical direction orthogonal to the circumferential direction. A filter coefficient determination device that determines the filter coefficient to be set to the digital filter to be
Calculating a directional basis on which the directivity of the sound output from the cylindrical array is based;
Filter coefficients of the cylindrical array using a filter basis, which is a function used to calculate the directional basis, and expansion coefficients obtained by giving a desired directivity to the directional basis A method of determining filter coefficients, including computing. A circular array, in which two or more speaker elements are arranged in the circumferential direction, is disposed in front of the speaker elements forming a cylindrical array in which two or more stages are stacked in the vertical direction orthogonal to the circumferential direction. The computer of the filter coefficient determination device for determining the filter coefficient to be set to the digital filter to be
Calculating a directional basis on which the directivity of the sound output from the cylindrical array is based;
Filter coefficients of the cylindrical array using a filter basis, which is a function used to calculate the directional basis, and expansion coefficients obtained by giving a desired directivity to the directional basis And a program for executing processing including. A cylindrical array in which a circular array in which two or more speaker elements are arranged in the circumferential direction is stacked in two or more stages in the vertical direction orthogonal to the circumferential direction;
And a digital filter disposed in front of the speaker elements constituting the cylindrical array;
Calculating a directional basis on which the directivity of the sound output from the cylindrical array is based,
Filter coefficients of the cylindrical array using a filter basis, which is a function used to calculate the directional basis, and expansion coefficients obtained by giving a desired directivity to the directional basis An acoustic system in which the filter coefficients determined by computing are set in the digital filter.

Images