Patent Translate Powered by EPO and Google Notice This translation is machine-generated. It cannot be guaranteed that it is intelligible, accurate, complete, reliable or fit for specific purposes. Critical decisions, such as commercially relevant or financial decisions, should not be based on machine-translation output. DESCRIPTION JP2000165984 [0001] BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sound amplification apparatus for amplifying an output sound using an electroacoustic equipment and to a method of improving the clarity of the apparatus. [0002] 2. Description of the Related Art Heretofore, as a method of improving the intelligibility in the case of sound amplification in the case of sound amplification, those described in the literature (School Technical Report EA 94-38) are known. FIG. 6 shows the configuration of a conventional sound amplification apparatus with improved clarity. In FIG. 6, 607 is a sound field, and there are N speakers 604 and M control points 606. An impulse response 605 between the speaker 604 and the control point 606 is obtained by measurement or prediction means. The coefficient calculator 611 calculates the coefficient 612 by the least square method using the target characteristic 609 in which the waveform with less reverberation is set, the transfer function setting unit 608 in which the impulse response 605 is set, and the divergence prevention coefficient 610. By setting this coefficient 612 in the sound field control filter 602 and connecting it to the speaker 604 via the amplifier 603, the input audio signal 601 can be clearly transmitted to the control point 606. [0003] In this way, even in the conventional sound amplification apparatus, the intelligibility of the loud 08-05-2019 1 speech can be temporarily improved. [0004] SUMMARY OF THE INVENTION However, in the above-described conventional method for improving the clarity in the sound amplification apparatus, when the reverberation time is long, the amount of calculation of the control filter required becomes enormous, and when the number of control points increases The number of sound field control filters also increases, and the number of impulse responses that need to be measured also increases. [0005] The present invention solves such conventional problems, and reduces the amount of calculation of a sound field control filter and reduces the number of sound field control filters while providing clear sound amplification apparatus and sound amplification apparatus It aims at providing the clarity improvement method. [0006] SUMMARY OF THE INVENTION According to the present invention, there is provided an acoustic loudspeaker and an intelligibility improvement method for calculating the coefficient of a sound field control filter, using singular value decomposition as the method of calculating coefficients of a sound field control filter. Among them, an element which is equal to or less than the reference value is set to 0 and then a means for calculating a coefficient, a means for setting a part of an impulse response between a speaker and a control point as a target characteristic, and a plurality of control points Control means, means for reducing the coefficient length and using only a part of the coefficients including the tap with the largest squared amplitude, and sound field control at a plurality of locations in a sound field such as a tunnel in which the same shape continues in one direction. The speaker and the control point are arranged at the same position at each control location, the sound field control filter coefficient is calculated at only one of a plurality of control locations, and the sound field control is performed at the other locations using this coefficient means And means for calculating the sound field control filter coefficient for controlling the plurality of control points at the same time, those having appropriately. As a result, the amount of calculation of the sound field control filter can be reduced, and clear sound can be performed while reducing the number of sound field control filters. 08-05-2019 2 [0007] According to the first aspect of the present invention, an input audio signal is amplified through a sound field control filter, and then, in the sound field, the speaker outputs a loud sound toward the control point. In a sound amplification apparatus that calculates a coefficient for a target characteristic based on an impulse response between a signal and a control point and sets the coefficient in a sound field control filter, singular value analysis is used as a method of calculating the coefficient of the sound field control filter. Of the elements of the singular value matrix for each frequency, the elements which are equal to or less than the reference value are set to 0 and then the apparatus is provided with means for calculating coefficients, and the calculation amount of the sound field control filter is reduced. This has the effect of being able to perform clear speech while reducing the number of sound field control filters. [0008] The invention according to claim 2 of the present invention is an acoustic loudspeaker according to claim 1, further comprising means for setting a part of an impulse response between the speaker and the control point as a target characteristic. It has the effect of reducing the amount of computation and performing clear sounding while reducing the number of sound field control filters. [0009] The invention according to claim 3 of the present invention is the sound amplification apparatus according to claim 1 or 2, wherein a plurality of control points are three-dimensionally arranged, wherein the calculation amount of the sound field control filter is reduced and the sound field control filter It has the effect of being able to carry out a clear loudspeaker while reducing the number of [0010] The invention according to claim 4 of the present invention shortens the length of the coefficient, and uses only a part of the coefficient including the tap where the square amplitude becomes maximum as the sound field control filter coefficient according to claim 3. It is an apparatus, and has the effect that the amount of calculation of the sound field control filter can be reduced, and clear speech can be performed while reducing the number of sound field control filters. [0011] The invention according to claim 5 of the present invention performs sound field control at a plurality of points in a sound field having the same shape continuously in one direction, such as a tunnel, and arranges speakers and control points at the same position at each control point, The 08-05-2019 3 acoustic loudspeaker according to any one of claims 1 to 4, wherein a sound field control filter coefficient is calculated at only one of a plurality of control points, and sound field control is performed at the other points using this coefficient. The present invention has the effect of being able to perform clear speech while reducing the amount of calculation of the sound field control filter and reducing the number of sound field control filters. [0012] The invention according to claim 6 of the present invention is the sound amplification apparatus according to claim 5, wherein sound field control filter coefficients for simultaneously controlling a plurality of control points are calculated, and the amount of calculation of the sound field control filter is reduced. And the effect of being able to perform clear speech while reducing the number of sound field control filters. [0013] According to the seventh aspect of the present invention, in the sound amplification apparatus, singular value decomposition is performed in the frequency domain as a method of calculating coefficients of the sound field control filter, and a threshold set in advance among respective elements of the singular value matrix The following element is an intelligibility improvement method of performing inverse matrix calculation after setting to 0, and has an effect that clear speech can be performed. [0014] The invention according to claim 8 of the present invention calculates the maximum value of each element of the singular value matrix at each frequency, and sets the threshold value based on the maximum value. It has the effect of being able to prevent divergence for each sound field control filter. [0015] The invention according to claim 9 of the present invention is the clarity improving method according to claim 8, wherein the threshold value is 40 dB smaller than the maximum value of each element of the singular value matrix at each frequency. It has the effect of preventing divergence for each filter. [0016] The invention according to claim 10 of the present invention is the clarity improving method according to any one of claims 7, 8 or 9, wherein a part of the impulse response between the 08-05-2019 4 speaker and the control point is used as the target characteristic. It has the effect of expanding the range in which clear speech can be performed to the periphery of the control point. [0017] The invention according to claim 11 of the present invention is the clarity improving method according to claim 10, wherein the impulse response between the speaker and the control point is set to 0 except for the part other than the direct sound as the target characteristic. It has the effect of expanding the range in which clear sound can be produced not only at the control point but also around the control point. [0018] The invention according to claim 12 of the present invention is the method for improving clarity according to claim 11, in which a plurality of control points are three-dimensionally arranged, and sterically expanding the range in which clear speech can be performed. It has an action. [0019] The invention according to claim 13 of the present invention is the method for improving clarity according to claim 12, wherein the control points are arranged on a three-dimensional lattice point whose one side is within 30 cm in length. It has the effect of three-dimensionally expanding the range that can be done. [0020] The invention according to claim 14 of the present invention performs sound field control by using only a part of coefficients including a tap at which the squared amplitude is maximum after coefficient calculation. This is a method of improving the clarity and has the effect of reducing the size of the required sound field control filter. [0021] The invention according to claim 15 of the present invention is the clarity improving method according to claim 14, wherein the coefficient length is reduced to 1/10 or less in comparison with the coefficient length before shortening. This has the effect of reducing the size of the filter. [0022] The invention according to claim 16 of the present invention performs sound field control at a plurality of locations in a sound field such as a tunnel in which the same cross-sectional shape is 08-05-2019 5 continuous in one direction, and arranges speakers and control points at the same location at each control location. The clarity improving method according to any one of claims 7 to 15, wherein the loudness can be clearly performed at a plurality of locations while reducing the size of the entire sound field control filter, and the number of required impulse responses. Have the effect of reducing [0023] The invention according to claim 17 of the present invention calculates the coefficient of the sound field control filter at only one of a plurality of control points, and executes the sound field control at other points using this coefficient. The clarity improving method according to the sixteenth aspect, which has an action of enabling clear sounding at a plurality of places while reducing the scale of the entire sound field control filter, and an action of reducing the number of necessary impulse responses. [0024] The invention according to claim 18 of the present invention is the clarity improving method according to claim 17, wherein the distance between the control points is 50 m or more, and has the effect of reducing the interference by other control points. [0025] The invention according to claim 19 of the present invention is the clarity improving method according to claim 16, wherein a sound field control filter coefficient for simultaneously controlling a plurality of control points is calculated, and the scale of the entire sound field control filter is reduced. However, it has the effect of being able to perform clear speech at multiple locations. [0026] The invention according to claim 20 of the present invention is the clarity improving method according to claim 19, wherein the filter coefficient is calculated using only an impulse response between a speaker and a control point in each control location, It has the effect of being able to perform clear sounding at multiple locations while reducing the overall scale of the sound field control filter. [0027] The invention according to claim 21 of the present invention is the method for improving clarity as claimed in claim 20, wherein the distance between the control points is 50 m or more, and has 08-05-2019 6 the effect of reducing interference by other control points. [0028] Hereinafter, embodiments of the present invention will be described with reference to FIGS. 1 to 5. (First Embodiment) FIG. 1 shows a first embodiment of the present invention. In FIG. 1, an input sound signal 101 is connected to N sound field control filters 102. The output of the sound field control filter 102 is connected to N speakers 104 in the sound field 107 via N amplifiers 103 respectively. In the sound field 107, the impulse response 105 between the N speakers 104 and the M control points 106 is measured, and based on this, the acoustic transfer function matrix 108 is calculated for each frequency. The acoustic transfer function matrix 108 is singular value decomposed by the singular value decomposition calculator 109. One of the outputs of the singular value decomposition calculator 109 is input to the inverse matrix calculator 111 via the singular value evaluator 110. Another output of the singular value decomposition calculator 109 is input to the inverse matrix calculator 111 as an input speech signal. The coefficient calculator 113 calculates the coefficient 114 based on the target characteristic 112 and the output of the inverse matrix calculator 111. 08-05-2019 7 The coefficient 114 is set to the sound field control filter 102. [0029] Next, the operation of the first embodiment will be described. In the sound field 107, N × M impulse responses 105 between N speakers 104 and M control points are obtained by measurement or simulation. The impulse response 105 is converted to the frequency domain by a method such as FFT, and the acoustic transfer function matrix 108 is calculated at each frequency. Here, the acoustic transfer function matrix 108 is represented by C (M, N). Each element of the matrix C is a complex number. The subscripts M and N mean that the matrix is an M-by-N matrix. The singular value decomposition calculator 109 performs singular value decomposition of the matrix C as shown in equation (1). C = UWVT (1) U (M, N): M × N column orthogonal matrix ((i = 1 to M) U U ik U jn = δkn, 1 <k <N, 1 <n <N) W (N, N): N × N diagonal matrix (diagonal components wj [j = 1 to N] nonnegative) VT (N, N): N × N orthogonal matrix V transpose [0030] When W obtained by the equation (1) is used as it is, if there is an element having a value close to 0 or almost 0 among the elements of W, the coefficient 114 of the sound field control filter corresponding to that element is There is a problem of diverging. 08-05-2019 8 Therefore, divergence is prevented for each sound field control filter in the singular value evaluator 110 by the following method. [0031] The singular value evaluator 110 finds the maximum value Wmax of each element of the singular value matrix W. Next, the ratio of each element to the maximum value Wmax is determined. If the ratio is less than or equal to the reference value, the value of the element wj is set to 0. The reference value may be in the range of −60 dB to −20 dB, and it is appropriate to set −40 dB. [0032] The inverse matrix calculator 111 calculates the inverse matrix of C according to equation (2). C-1 = V [diag (1 / wj)] UT .. Formula (2) diag (1 / wj): N × N diagonal matrix with an element of 1 / wj [j = 1 to N] [0033] Impulse responses with few reverberation components are set as targets at M control points, converted into frequency domains, and a matrix B (M) is set as a target characteristic 112 at each frequency. 08-05-2019 9 The coefficient calculator 113 calculates the solution X (N) according to equation (3). X = C-1B ・ ・ ・ Formula (3) [0034] X is obtained for each frequency, the frequency characteristics of the N sound field control filters are obtained, and each is converted to the time domain by a method such as inverse FFT to obtain coefficients 114. By setting this coefficient in the sound field control filter 102, the control point 106 can be clearly loudened. Thus, clear sound can be performed while preventing divergence for each sound field control filter. [0035] Second Embodiment FIG. 2 shows a second embodiment of the present invention. In FIG. 2, the input sound signal 201 is connected to N sound field control filters 202. The output of the sound field control filter 202 is connected to N speakers 204 in the sound field 207 via N amplifiers 203 respectively. In the sound field 207, impulse responses 205 between the N speakers 204 and the M control points 206 are measured. Based on the impulse response 205, the acoustic transfer function matrix 208 is calculated for each frequency. The impulse response 205 is also input to the target characteristic setting unit 215. The acoustic transfer function matrix 208 is input to the singular value decomposition calculator 209. The singular value decomposition calculator 209 performs singular value decomposition of the acoustic transfer function matrix 208. One of the outputs of the singular value decomposition calculator 209 is input to the inverse matrix calculator 211 via the singular value evaluator 210. Another output of the singular value decomposition calculator 209 is input to the inverse matrix calculator 211. The coefficient calculator 213 calculates the coefficient 214 based on the target characteristic 212 which is the output of the target characteristic setting unit 215 and the output of the inverse matrix calculator 211. The coefficient 214 is set to the sound field control filter 202. [0036] Next, the operation of the second embodiment will be described. In the sound field 207, N × M impulse responses 205 between N speakers 204 and M control points are obtained by 08-05-2019 10 measurement or simulation. M control points are arranged on a straight line. The distance d between the control points is preferably 30 cm or less, and in practice it is appropriate to be 10 cm. The impulse response 205 is converted to the frequency domain by a method such as FFT, and the acoustic transfer function matrix 208 is calculated at each frequency. The acoustic transfer function matrix 208 is an input of the singular value decomposition calculator 209. Here, the acoustic transfer function matrix 208 is represented by C (M, N). Each element of the matrix C is a complex number. The subscripts M and N mean that the matrix is an M-by-N matrix. The singular value decomposition calculator 209 performs singular value decomposition of the matrix C as shown in equation (1). [0037] When W obtained by the equation (1) is used as it is, if there is an element having a value close to 0 or almost 0 among the elements of W, the coefficient 214 of the sound field control filter corresponding to that element is There is a problem of diverging. Therefore, the singular value evaluator 210 prevents divergence for each sound field control filter by the following method. [0038] The singular value evaluator 210 finds the maximum value Wmax of each element of the singular value matrix W. Next, the ratio of each element to the maximum value Wmax is determined. If the ratio is less than or equal to the reference value, the value of the element wj is set to 0. The reference value may be in the range of −60 dB to −20 dB, and it is appropriate to set −40 dB. [0039] The target characteristic setting unit 215 receives the impulse response 205 and calculates the direct sound part of the impulse response 205. The calculation method of the direct sound part of the impulse response is performed as follows. One of the N speakers 204 is selected, and taps that maximize the squared amplitude of the impulse response between that speaker and the M listening points are respectively determined. The window function is multiplied around the tap, and the range of tap ± T where the squared amplitude becomes maximum is set as the direct sound part of the impulse response. The shape of the window function may be a rectangular window or a Hanning window. The optimum magnitude of T is within the range in which the 08-05-2019 11 direct sound response of the speaker used is contained. Generally, it is in the range of 10 to 50 taps. The direct sound parts of the M impulse responses thus obtained are converted to the frequency domain by a method such as FFT, and the matrix B (M) is set as the target characteristic 212 at each frequency. The inverse matrix calculator 211 calculates the inverse matrix of C according to equation (2). [0040] The coefficient calculator 213 calculates the solution as a solution X (N) according to equation (3). By obtaining X for each frequency, the frequency characteristics of the N sound field control filters are calculated. The frequency characteristics of the respective sound field control filters are converted to the time domain by a method such as inverse FFT to obtain coefficients 214. By setting this coefficient in the sound field control filter 202, it is possible to clearly louden at the control point 206 and its surroundings. Thus, clear sound can be performed while preventing divergence for each sound field control filter. [0041] Third Embodiment FIG. 3 shows a third embodiment of the present invention. In FIG. 3, the input sound signal 301 is connected to N sound field control filters 302. The output of the sound field control filter 302 is connected to the N speakers 304 in the sound field 307 via N amplifiers 303 respectively. In the sound field 307, impulse responses 305 between the N speakers 304 and the M control points 306 are measured. Based on the impulse response 305, an acoustic transfer function matrix 308 is calculated for each frequency. The impulse response 305 is also input to the target characteristic setting unit 315. The acoustic transfer function matrix 308 is input to the singular value decomposition calculator 309. The singular value decomposition calculator 309 performs singular value decomposition of the acoustic transfer function matrix 308. One of the outputs of the singular value decomposition calculator 309 is input to the inverse matrix calculator 311 via the singular value evaluator 310. Another output of the singular value decomposition calculator 309 is input to the inverse matrix calculator 311. The coefficient calculator 313 calculates the coefficient 314 based on the target characteristic 312 which is the output of the target characteristic setting unit 315 and the output of the inverse matrix calculator 311. The calculated coefficient 314 finds a tap with the maximum squared amplitude at the maximum value detector 316, and a window function is multiplied by the coefficient length changer 317 around that tap to obtain the range of taps ± T where the squared amplitude becomes maximum. Is calculated as a new coefficient, and is set in the sound field control filter 302. 08-05-2019 12 [0042] Next, the operation of the third embodiment will be described. In the sound field 307, N × M impulse responses 305 between N speakers 304 and M control points 306 are obtained by measurement or simulation. The M control points 306 are three-dimensionally arranged. The distance d between the control points 306 is preferably 30 cm or less, and in practice, 10 cm to 20 cm is appropriate. In the practical arrangement of control points 306, 27 control points are arranged at a distance d of 10 cm in a cube having a side length of 20 cm, which is 1.5 m in height from the floor surface. [0043] The impulse response 305 is converted to the frequency domain by a method such as FFT, and the acoustic transfer function matrix 308 is calculated at each frequency. The acoustic transfer function matrix 308 is an input of the singular value decomposition calculator 309. Here, the acoustic transfer function matrix 308 is represented by C (M, N). Each element of the matrix C is a complex number. The subscripts M and N mean that the matrix is an M-by-N matrix. The singular value decomposition calculator 309 performs singular value decomposition of the matrix C as shown in equation (1). [0044] When W obtained by the equation (1) is used as it is, if there is an element having a value close to 0 or almost 0 among the elements of W, the coefficient 314 of the sound field control filter corresponding to that element is There is a problem of diverging. Therefore, divergence is prevented for each sound field control filter in the singular value evaluator 310 by the following method. [0045] The singular value evaluator 310 finds the maximum value Wmax of each element of the singular value matrix W. Next, the ratio of each element to the maximum value Wmax is 08-05-2019 13 determined. If the ratio is less than or equal to the reference value, the value of the element wj is set to 0. The reference value may be in the range of −60 dB to −20 dB, and it is appropriate to set −40 dB. [0046] The target characteristic setting unit 315 receives the impulse response 305 and calculates the direct sound part of the impulse response 305. The calculation method of the direct sound part of the impulse response 305 is performed as follows. One of the N speakers 304 is selected, and taps that maximize the squared amplitude of the impulse response between that speaker and the M listening points are respectively determined. The window function is multiplied around the tap, and the range of tap ± T where the squared amplitude becomes maximum is set as the direct sound part of the impulse response. The shape of the window function may be a rectangular window or a Hanning window. The optimum magnitude of T is within the range in which the direct sound response of the speaker used is contained. Generally, it is in the range of 10 to 50 taps. The direct sound parts of the M impulse responses thus obtained are converted to the frequency domain by a method such as FFT, and the matrix B (M) is set as the target characteristic 312 at each frequency. The inverse matrix calculator 311 calculates the inverse matrix of C according to equation (2). [0047] The coefficient calculator 313 calculates as a solution X (N) according to equation (3). X for each frequency is determined, the frequency characteristics of the N sound field control filters are determined, and each is converted to the time domain by a method such as inverse FFT to obtain N reduced pre-coefficients. The maximum value detector 316 finds taps that maximize the squared amplitude of the pre-shortening coefficient. The coefficient length changer 317 multiplies the window function around the tap where the squared amplitude of the preshortening coefficient is maximized, and sets the range of taps ± T where the squared amplitude is maximized as a new coefficient. The shape of the window function may be a rectangular window or a Hanning window. Although it is appropriate to set the size of T to 1/10 of the length of the pre-reduction coefficient, it may be further shortened. [0048] By setting the coefficients 314 thus calculated in the sound field control filter 302, it is possible to clearly louden at the control point 306 and its surroundings, and to perform clear loudness 08-05-2019 14 while reducing the scale of the sound field control filter. Can. [0049] (Fourth Embodiment) FIG. 4 shows a fourth embodiment of the present invention. In FIG. 4, an input sound signal 401 is connected to N sound field control filters 402. The output of the sound field control filter 402 is distributed to the N speakers 404 at the control point A 416 and the N speakers 418 at the control point B 417 in the sound field 407 through N amplifiers 402 respectively. The sound field 407 is a sound field such as a tunnel in which the same cross-sectional shape is continuous in one direction. In the sound field 407, an impulse response 405 between the speaker 404 and the control point 406 is defined at the control point A416, and an impulse response 419 between the speaker 418 and the control point 420 is defined at the control point B417. Based on the impulse response 405, an acoustic transfer function matrix 408 is calculated for each frequency. The impulse response 405 is also input to the target characteristic setting unit 415. The acoustic transfer function matrix 408 is input to the singular value decomposition calculator 409. The singular value decomposition calculator 409 performs singular value decomposition of the acoustic transfer function matrix 408. One of the outputs of the singular value decomposition calculator 409 is input to the inverse matrix calculator 411 via the singular value evaluator 410. Another output of the singular value decomposition calculator 409 is input to the inverse matrix calculator 411. The coefficient calculator 413 calculates the coefficient 414 based on the target characteristic 412 which is the output of the target characteristic setting unit 415 and the output of the inverse matrix calculator 411. The coefficient 414 is set to the sound field control filter 402. [0050] Next, the operation of the fourth embodiment will be described. In the sound field 407, N × M impulse responses 405 between the N speakers 404 at the control point A 416 and the M control points 406 are obtained by measurement or simulation. Here, N × M impulse responses 419 between the N speakers 418 of the control point B 417 and the M control points 420 are not used. [0051] 08-05-2019 15 The sound field 407 is a shape such as a tunnel in which the same cross-sectional shape is continuous in one direction. Therefore, the cross-sectional shape in control location A416 and control location B417 is the same. The larger the distance L between the control point A 416 and the control point B 417, the better, and it is necessary that the distance L be 50 m or more. [0052] At the control point A 416, the M control points 406 are three-dimensionally arranged. The distance d between the control points 406 is preferably 30 cm or less, and in practice, 10 cm to 20 cm is appropriate. In the practical arrangement of control points 406, 27 control points are arranged at a distance d of 10 cm in a cube having a side length of 20 cm, which is 1.5 m in height from the floor surface. In addition, the speaker 404 is disposed at the control point A 416. The speakers 404 are disposed so as to cover the control points 406, respectively, and are disposed close to each other so that the distance between the speakers is as narrow as possible. [0053] At the control point B 417, the control point 420 is arranged the same as the control point 406 at the control point A 416. Further, the speaker 418 is also arranged the same as the speaker 404 at the control point A 416. As described above, arranging the loudspeaker and the control point in the same manner at each of the control point A 416 and the control point B 417 has an effect of enhancing the correlation between the impulse response 405 at the control point A 416 and the impulse response 419 at the control point B 417. [0054] The N × M impulse responses 405 between the N speakers 404 at the control point A 416 and the M control points 406 are respectively converted to the frequency domain by a method such as FFT, and the acoustic transfer function matrix 408 is calculated at each frequency. Do. The acoustic transfer function 408 is an input of the singular value decomposition calculator 409. Here, the acoustic transfer function matrix 408 is represented by C (M, N). Each element of the matrix C is a complex number. The subscripts M and N mean that the matrix is an M-by-N matrix. The singular value decomposition calculator 409 performs singular value decomposition of the matrix C as shown in equation (1). 08-05-2019 16 [0055] When W obtained by the equation (1) is used as it is, if there is an element having a value close to 0 or almost 0 among the elements of W, the coefficient 414 of the sound field control filter corresponding to that element is There is a problem of diverging. Therefore, divergence is prevented for each sound field control filter in the singular value evaluator 410 by the following method. [0056] The singular value evaluator 410 finds the maximum value Wmax of each element of the singular value matrix W. Next, the ratio of each element to the maximum value Wmax is determined. If the ratio is less than or equal to the reference value, the value of the element wj is set to 0. The reference value may be in the range of −60 dB to −20 dB, and it is appropriate to set −40 dB. [0057] The target characteristic setting unit 415 receives the impulse response 405 and calculates the direct sound part of the impulse response 405. The method of calculating the direct sound portion of the impulse response 405 is as follows. One of the N speakers 404 is selected, and taps that maximize the squared amplitude of the impulse response between that speaker and the M listening points are respectively determined. The window function is multiplied around the tap, and the range of tap ± T where the squared amplitude becomes maximum is set as the direct sound part of the impulse response. The shape of the window function may be a rectangular window or a Hanning window. The optimum magnitude of T is within the range in which the direct sound response of the speaker used is contained. Generally, it is in the range of 10 to 50 taps. The direct sound parts of the M impulse responses thus obtained are converted to the frequency domain by a method such as FFT, and the matrix B (M) is set as the target characteristic 412 at each frequency. The inverse matrix calculator 411 calculates the inverse matrix of C according to equation (2). [0058] 08-05-2019 17 The coefficient calculator 413 calculates the solution as a solution X (N) according to equation (3). X is obtained for each frequency, the frequency characteristics of the N sound field control filters are obtained, and N coefficients 414 are obtained by converting each to the time domain by a method such as inverse FFT. By setting the coefficients 414 thus calculated in the sound field control filter 402, the control point 406 of the control point A 416 and its periphery, and the control point 420 of the control point B 417 and its two peripheral points can be clearly amplified. It is possible to perform clear sounding at a plurality of locations while reducing the size of the entire sound field control filter. Although the number of control points is two here, it is apparent that the same result can be obtained even if the number of control points is two or more. [0059] (Fifth Embodiment) FIG. 5 shows a fifth embodiment of the present invention. In FIG. 5, an input sound signal 501 is connected to N sound field control filters 502. The output of the sound field control filter 502 is distributed to the N speakers 504 at the control point A 516 and the N speakers 518 at the control point B 517 in the sound field 507 through N amplifiers 503 respectively. The sound field 507 is a sound field such as a tunnel in which the same crosssectional shape is continuous in one direction. In the sound field 507, an impulse response 505 between the speaker 504 and the control point 506 is defined at the control point A516, and an impulse response 519 between the speaker 518 and the control point 520 is defined at the control point B517. Further, an impulse response 521 and an impulse response 522 are defined between the control point A 516 and the control point B 517. Based on the impulse response 505 and the impulse response 519, an acoustic transfer function matrix 508 is calculated for each frequency. The impulse response 505 and the impulse response 522 are also input to the target characteristic setting unit 515. The acoustic transfer function matrix 508 is input to the singular value decomposition calculator 509. The singular value decomposition calculator 509 performs singular value decomposition of the acoustic transfer function matrix 508. One of the outputs of the singular value decomposition calculator 509 is input to the inverse matrix calculator 511 through the singular value evaluator 510. Another output of the singular value decomposition calculator 509 is input to the inverse matrix calculator 511. The coefficient calculator 513 calculates the coefficient 514 based on the target characteristic 512 which is the output of the target characteristic setting unit 515 and the output of the inverse matrix calculator 511. The coefficient 514 is set in the sound field control filter 502. [0060] 08-05-2019 18 Next, the operation of the fifth embodiment will be described. In the sound field 507, N × M impulse responses 505 between the N speakers 504 of the control point A 516 and the M control points 506 are obtained by measurement or simulation. Furthermore, N × M impulse responses 519 between the N speakers 518 of the control point B 517 and the M control points 520 are determined by measurement or simulation. Here, also, the impulse response 521 and the impulse response 522 between the control point A 516 and the control point B 517 are not used. [0061] The sound field 507 is a shape such as a tunnel in which the same cross-sectional shape is continuous in one direction. Therefore, the cross-sectional shape in control location A516 and control location B517 is the same. It is preferable that the distance L between the control point A 516 and the control point B 517 is as wide as possible, and 50 m or more is necessary, and in practice, it is preferably 100 m or more. [0062] At control point A 516, M control points 506 are arranged three-dimensionally. The distance d between the control points 506 is preferably 30 cm or less, and in practice, 10 cm to 20 cm is appropriate. In the practical arrangement of control points 506, 27 control points are arranged at a distance d of 10 cm in a cube having a side length of 20 cm, which is 1.5 m above the floor surface. Further, the speaker 504 is disposed at the control point A 516. The speakers 504 are disposed so as to cover the control points 506, respectively, and are disposed close to each other such that the distance between the speakers is as narrow as possible. [0063] At the control point B 517, the control point 520 is arranged the same as the control point 506 at the control point A 516. Further, the speaker 518 is also arranged the same as the speaker 504 at the control point A 516. As described above, arranging the loudspeaker and the control point in the same manner in each of the control point A 516 and the control point B 517 has an effect of enhancing the correlation between the impulse response 505 in the control point A 516 08-05-2019 19 and the impulse response 519 in the control point B 517. [0064] The acoustic transfer function matrix A is calculated by converting the N × M impulse responses 505 between the N speakers 504 of the control location A 516 and the M control points 506 into the frequency domain by a method such as FFT. Further, the acoustic transfer function matrix B is calculated by converting the N × M impulse responses 519 between the N speakers 518 at the control location B 517 and the M control points 520 into the frequency domain by a method such as FFT. . The acoustic transfer function matrices A (N, M) and B (N, M) are complex matrices of N rows and M columns, respectively. Here, the acoustic transfer function matrix C (N, 2M) is calculated by combining A and B as shown in equation (4). [0065] The matrix C obtained in this manner is referred to as an acoustic transfer function matrix 508. The acoustic transfer function matrix 508 is input to the singular value decomposition calculator 509. The singular value decomposition calculator 509 performs singular value decomposition of the matrix C as expressed by equation (1). [0066] When W obtained by the equation (1) is used as it is, if there is an element having a value close to 0 or almost 0 among the elements of W, the coefficient 514 of the sound field control filter corresponding to that element is There is a problem of diverging. Therefore, divergence is prevented for each sound field control filter in the singular value evaluator 510 by the following method. [0067] The singular value evaluator 510 obtains the maximum value Wmax of each element of the singular value matrix W. Next, the ratio of each element to the maximum value Wmax is determined. If the ratio is less than or equal to the reference value, the value of the element wj is 08-05-2019 20 set to 0. The reference value may be in the range of −60 dB to −20 dB, and it is appropriate to set −40 dB. [0068] The target characteristic setting unit 515 receives the impulse response 505 and the impulse response 522, and calculates the direct sound part of the impulse response 505 and the impulse response 522. The calculation method of the direct sound part of the impulse response is performed as follows. One of the N speakers 504 is selected, and taps that maximize the squared amplitude of the impulse response between that speaker and the M listening points are respectively determined. The window function is multiplied around the tap, and the range of tap ± T where the squared amplitude becomes maximum is set as the direct sound part of the impulse response. The shape of the window function may be a rectangular window or a Hanning window. The optimum magnitude of T is within the range in which the direct sound response of the speaker used is contained. Generally, it is in the range of 10 to 50 inches. The direct sound parts of the M impulse responses thus obtained are converted to the frequency domain by a method such as FFT, and the matrix B (2M) is set as the target characteristic 512 at each frequency. The inverse matrix calculator 511 calculates the inverse matrix of C according to equation (2). [0069] The coefficient calculator 513 calculates the solution as a solution X (N) according to equation (3). For each frequency, X is obtained, the frequency characteristics of the N sound field control filters are obtained, and each is converted to the time domain by a method such as inverse FFT to obtain N coefficients 514. By setting the coefficient 514 thus calculated in the sound field control filter 502, the control point 506 of the control point A 516 and its surroundings, and the control point 520 of the control point B 517 and its two surrounding points can be clearly amplified. It is possible to perform clear sounding at a plurality of locations while reducing the size of the entire sound field control filter. Although the number of control points is two here, it is apparent that the same result can be obtained even if the number of control points is two or more. [0070] As described above, the present invention performs singular value decomposition in the frequency domain as a method of calculating coefficients of the sound field control filter, and among the elements of the singular value matrix, elements smaller than a preset threshold value. 08-05-2019 21 Is a sound amplification apparatus that performs inverse matrix calculation after setting 0, and a method of improving the clarity thereof, and can perform clear sound while reducing the number and calculation amount of sound field control filters. 08-05-2019 22

1/--страниц