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 JP6313517 Abstract: PROBLEM TO BE SOLVED: To provide a filter coefficient calculation device and a hearing aid which can easily determine the coefficients of a plurality of digital filters related to negative feedback control and realize optimization of the suppression amount in a short time. A filter coefficient calculation device 1 for determining coefficients of a plurality of digital filters related to negative feedback control, the acquisition unit 11 acquiring transfer function data E '(jω) based on a transfer function E (s) , A population generation unit 12 that generates a population including a plurality of reference data, an optimization unit 13 that repeats selection and updating a plurality of times using a genetic algorithm, and an output unit 14, and the optimization unit 13 Generates evaluation data R (ω) including a child data generation unit 13a that generates a plurality of child data ti, transfer function data E ′ (jω), and child data ti, and generates evaluation data R (ω) It is characterized in that it includes an evaluation unit 13 b that performs evaluation, and repeats generation and evaluation, and an update unit 13 c. [Selected figure] Figure 1 Filter coefficient calculation device and hearing aid [0001] The present invention relates to a filter coefficient calculation device that determines coefficients of a plurality of digital filters related to negative feedback control, and a hearing aid. [0002] Heretofore, as an application of circuit technology using negative feedback control including a 16-04-2019 1 digital filter, for example, a muddy noise reduction device as disclosed in Patent Document 1 has been proposed. Also, for example, Patent Document 2 discloses a method of determining coefficients of a digital filter. [0003] The mumbling noise reduction device disclosed in Patent Document 1 includes a sound output unit having an opening formed therein, a sound collecting unit having an opening formed therein, and a sound conducting unit communicating the opening with a space in the ear canal. And a negative feedback unit that performs negative feedback from the sound collection unit to the sound output unit. A common space, one end of which faces the opening and the other end of which faces the space in the ear canal, is formed inside the sound guiding portion, and sound propagates in both directions via the common space. The common space relates to the length L (mm) and the cross-sectional area S (mm <2>), and within the range of 3 ≦ L ≦ 8, S ≧ a1 · L + b1 and S ≦ a2 · L + b2, where a1 = 0 It is set to satisfy the relationship of .19 (mm), b1 = 0.57 (mm <2>), a2 = 0.65 (mm), b2 = 0.18 (mm <2>). [0004] The coefficient determination method of the digital filter disclosed in Patent Document 2 generates N sets (individuals) obtained by quantizing each of n coefficient values before quantization according to an arbitrary quantization method by the initial group generation unit And the genetic algorithm executor repeatedly applies the genetic algorithm to the initial population, and the genetic algorithm causes the n individual coefficient values to be reduced so as to reduce degradation from the desired frequency characteristics. It quantizes according to a predetermined quantization method to obtain coefficient data of a finite word length. [0005] JP, 2017-11383, A JP, 09-046178, A [0006] Here, in the design of a circuit used for negative feedback control, among various limitations, it 16-04-2019 2 may be necessary to derive an optimal characteristic. In particular, in order to realize an optimal amount of suppression in negative feedback control, a plurality of digital filters may be provided, and it is necessary to set coefficients for determining the frequency characteristics of each digital filter. Also, with slight differences in the mounted environment (use conditions), the coefficients of the digital filter must be changed to achieve the optimum amount of suppression. For this reason, it is not easy to determine the coefficient of the digital filter, and it is necessary to spend a great deal of time to realize the optimization of the suppression amount. [0007] In this regard, the technology disclosed in Patent Document 1 does not describe the determination of the coefficient of the digital filter, and the above-described situation can not be solved. Further, in the technology disclosed in Patent Document 2, a digital filter coefficient is determined using a genetic algorithm based on bit operation. For this reason, it is unsuitable to determine the coefficients of a plurality of digital filters used for the above-mentioned negative feedback control, and the above-mentioned situation can not be solved. From the abovementioned circumstances, it is desired that coefficients of a plurality of digital filters can be easily determined and optimization of the suppression amount can be realized in a short time. [0008] Therefore, the present invention has been made in view of the above-mentioned problems, and the purpose of the present invention is to easily determine the coefficients of a plurality of digital filters related to negative feedback control and shorten the optimization of the suppression amount. It is an object of the present invention to provide a filter coefficient calculation device that can be realized in time and a hearing aid. [0009] A filter coefficient calculation device according to a first aspect of the present invention is directed to negative feedback control connected between the receiver and the microphone based on a transfer function calculated using a signal input to the receiver and a signal output from the microphone. A filter coefficient calculation device that determines coefficients of a plurality of digital filters, including an acquisition unit that acquires transfer function data based on the 16-04-2019 3 transfer function, and a plurality of reference data set based on initial information acquired in advance. Selecting a part of the reference data of the population using a population generation unit for generating a population and a genetic algorithm; updating the selected part of the reference data; The optimization unit that repeats the update a plurality of times, first reference data is selected from the population, and the plurality of digital filters are selected based on the first reference data. A child data generation unit that generates a plurality of child data based on the selected part of the reference data; the transfer function data; Evaluation data including child data is generated, the evaluation data is evaluated, an evaluation unit which repeats the generation and the evaluation, and a part of the child data of the plurality of child data based on a result of the evaluation And a updating unit that selects and updates the selected part of the reference data. [0010] A filter coefficient calculation device according to a second aspect of the present invention is the filter coefficient calculation device according to the first aspect, wherein the evaluation data includes an overall gain value, and the evaluation unit evaluates the entire when the evaluation data satisfies a preset condition. The gain value is updated, and the evaluation data is evaluated if the condition is not satisfied, and the updating unit is configured to select the selected partial data and the selected partial data. And updating the entire gain value corresponding to the selected part as the reference data, and the coefficients of the output plurality of digital filters are corresponding to the first reference data. It is characterized by having a gain value. [0011] The filter coefficient calculation device according to the third invention is the filter coefficient calculation device according to the second invention, wherein the reference data and the child data have a center frequency, a quality coefficient, and a gain value, and the evaluation unit ] The said evaluation data (R ((omega))) shown to these is produced | generated, It is characterized by the above-mentioned. Here, L (s) is shown by the following [Equation 14] as a loop transfer function. Here, Gall represents the entire gain value, E ′ (s) represents the transfer function data, and H (s) is represented by the following Equation 13 as a transfer function of the controller. 16-04-2019 4 Here, fcik, Qik, and Gik respectively indicate the center frequency, the quality factor, and the gain value of the child data corresponding to the evaluation data. [0012] The filter coefficient calculation device according to a fourth aspect of the invention is the filter coefficient calculation device according to the third aspect, wherein the evaluation unit determines the evaluation data with the following [Equation 16] as the condition, and updates the evaluation data based on the result of the determination Or characterized by said evaluation. Here, fd1 indicates the first frequency corresponding to the above condition, plow indicates the upper limit value of the first frequency or less, and pup indicates the upper limit value of the first frequency or more. [0013] The filter coefficient calculation device according to a fifth aspect of the present invention is the filter coefficient calculation device according to the fourth aspect, wherein the evaluation unit determines the evaluation data using the equation 16 and the following equation 17 as conditions. As a result, the evaluation data is updated or evaluated. Here, fd2 indicates the second frequency corresponding to the above condition, and nlow indicates the lower limit value not higher than the second frequency. [0014] In the filter coefficient calculation device according to the sixth invention, in any one of the third invention to the fifth invention, the evaluation unit refers to the evaluation function (E) represented by the following [Equation 18], and the evaluation data The above evaluation. Here, fi indicates an evaluation frequency, Rtg (fi) indicates a target value at the evaluation frequency, C1 indicates a first weighting factor, and C2 indicates a second weighting factor. [0015] In the filter coefficient calculation device according to the seventh invention, in any one of the 16-04-2019 5 third invention to the sixth invention, the acquisition unit refers to the following [Equation 2], the transfer function data (E ′) based on the transfer function. (Jω)) is obtained. Here, E (jω) represents the transfer function, fl represents a correction frequency to be corrected, and β represents a phase gradient. [0016] The filter coefficient calculation device according to an eighth aspect of the invention is the filter coefficient calculation device according to any one of the third to seventh aspects, wherein the population generation unit includes a central frequency distribution (fc) represented by the following [Equation 3] of the initial information. And setting the center frequency based on. Here, ε ini indicates a random number, and f cup indicates an upper limit frequency. [0017] The filter coefficient calculation device according to a ninth aspect of the present invention is the filter coefficient calculation device according to any of the third aspect to the eighth aspect, wherein the population generation unit has a quality factor distribution (Q) represented by the following [Equation 4] of the initial information. And setting the quality value based on a gain value distribution (G) represented by the following [Equation 5] of the initial information. Here, Qup indicates an upper limit quality factor, Gup indicates an upper limit gain value, and εini indicates a random number. [0018] A filter coefficient calculation device according to a tenth aspect of the present invention is connected between the transmitter and the receiver based on a signal input to the transmitter and a transfer function calculated using the signal output from the receiver. A filter coefficient calculation device for determining coefficients of a plurality of digital filters relating to negative feedback control, the acquisition unit acquiring transfer function data based on the transfer function, and a plurality of sets set based on initial information acquired in advance. Selecting a part of the reference data of the population using a population generation unit for generating a population including reference data, and using a genetic algorithm, and updating the selected part of the reference data And an optimization unit that repeats the selection and the update, and selects a first reference data from the population, and outputs coefficients of the plurality of digital filters based on the first reference data. A child data generation unit that generates a plurality of child data based on the selected part of the reference data, a child data generation 16-04-2019 6 unit, the transfer function data, and the child data Generating evaluation data, evaluating the evaluation data, selecting an evaluation unit that repeats the generation and the evaluation, and a part of the child data among the plurality of child data based on the result of the evaluation; And updating the update unit as the selected part of the reference data. [0019] A hearing aid according to an eleventh aspect of the present invention comprises an external microphone that converts sound transmitted from an external space into an electrical signal, a microphone for the ear canal that converts sound transmitted from the inside of the ear canal into an electrical signal, and the electric signal converted by the external microphone A hearing aid processing unit that performs hearing aid processing according to a user, a receiver that outputs a sound converted from an electrical signal, a plurality of digital filters related to negative feedback control disposed between the receiver and the ear canal microphone, A controller having a filter coefficient calculation unit that determines coefficients of the plurality of digital filters, the filter coefficient calculation unit using a signal input to the receiver and a signal output from the ear canal microphone An acquisition unit for acquiring the calculated transfer function and transfer function data based on the transfer function; A population generation unit for generating a population including a plurality of reference data set based on the acquired initial information; and using the genetic algorithm to select the reference data of a part of the population, The optimization unit updates the selected part of the reference data, repeats the selection and the update, and selects the first reference data from the population, and the plurality of digital filters based on the first reference data. A child data generation unit that generates a plurality of child data based on the selected part of the reference data, and the transfer function data. An evaluation unit generating evaluation data including the child data, evaluating the evaluation data, repeating the generation and the evaluation, and a part of the plurality of child data based on a result of the evaluation Select data, and having a an update unit for the update as the reference data of a portion of the selected. [0020] According to the first to eleventh inventions, the evaluation unit evaluates evaluation data including transfer function data and child data using a genetic algorithm. Also, transfer function data is obtained based on the transfer function of the environment using the digital filter. 16-04-2019 7 For this reason, the evaluation unit can carry out the evaluation in consideration of the difference in the environment in which the digital filter is used. This makes it possible to easily determine the coefficients of the plurality of digital filters and realize optimization of the suppression amount in a short time. [0021] Further, according to the second invention, the plurality of evaluation data includes the entire gain value, and the evaluation unit updates the entire gain value when the evaluation data satisfies the condition. That is, an optimal overall gain value can be set for each child data generated by the genetic algorithm. Therefore, by using the entire gain as a coefficient of the digital filter, the entire suppression amount can be adjusted, and the accuracy of the negative feedback control can be improved. [0022] Further, according to the third aspect of the present invention, the evaluation unit generates evaluation data represented by [Equation 15]. At this time, the evaluation data has the center frequency of the child data, the quality factor, the gain value, and the overall gain value, and corresponds to the suppression amounts of the plurality of digital filters using the respective values as coefficients. For this reason, the evaluation unit can carry out the evaluation based on the actual suppression amount. This makes it possible to determine the coefficients of the plurality of digital filters with high accuracy. [0023] Further, according to the fourth aspect of the invention, the evaluation unit determines the evaluation data under the condition of [Equation 16], and updates or evaluates the evaluation data based on the result of the determination. [Equation 16] includes an expression for comparing the value of the evaluation data with respect to the upper limit value of the first frequency or lower and the upper limit value of the first frequency or higher. For this reason, the upper limit value of positive feedback in each evaluation data (suppression amount) can be set arbitrarily. This makes it possible to suppress unintended signal amplification when the determined digital filter coefficients are used. 16-04-2019 8 [0024] According to the fifth aspect of the invention, the evaluation unit discriminates the evaluation data on the condition of [Equation 16] and [Equation 17], and updates or evaluates the evaluation data based on the result of the discrimination. [Equation 17] is an expression for comparing the value of the evaluation data with respect to the lower limit value of the second frequency or lower. For this reason, when using a digital filter, it is possible to arbitrarily set a frequency band that does not require negative feedback control. This makes it possible to suppress the consumption of power. [0025] Further, according to the sixth invention, the evaluation unit evaluates the evaluation data with reference to the evaluation function represented by [Equation 18]. [Equation 18] can indicate the difference between the target value at a plurality of evaluation frequencies and the evaluation data as a unique value. Therefore, it is possible to easily select some child data among the plurality of child data based on the evaluation result, and to easily converge the child data to the target value. This makes it possible to further shorten the time required to optimize the suppression amount. [0026] Further, according to the seventh invention, the acquiring unit acquires transfer function data based on the transfer function with reference to [Equation 2]. [Equation 2] is an expression that flattens the amplitude characteristic and rotates the phase characteristic in the transfer function below the correction frequency. For this reason, it is possible to determine the filter coefficient in consideration of the change of the transfer function. This makes it possible to increase the stability of the suppression amount even when the transfer function in the environment using the digital filter changes. [0027] Further, according to the eighth invention, the population generation unit sets the center 16-04-2019 9 frequency based on the center frequency distribution. That is, the range of possible center frequencies of reference data can be limited in advance. For this reason, when using a genetic algorithm, the reference data to be updated can be reliably converged. This makes it possible to further shorten the time required to optimize the suppression amount. [0028] Further, according to the ninth aspect, the quality factor and the gain value are set based on the quality factor distribution and the gain value distribution. Here, the frequency characteristics of the peaking filter included in the digital filter are determined by the frequency of the peak, the quality factor, and the gain value. Therefore, by limiting the upper limit of the quality factor and the gain value in advance, it is possible to eliminate the factor that produces the extreme peak. This makes it possible to improve the stability of the suppression amount when using a peaking filter as the digital filter. [0029] FIG. 1 is a schematic view showing an example of the configuration of a filter coefficient calculation device according to the embodiment. FIG. 2 (a) is a graph showing an example of the frequency characteristic of the transfer function, and FIG. 2 (b) is a graph showing an example of the phase characteristic of the transfer function. FIG. 3 (a) shows an example of the equivalent circuit of the negative feedback controller, FIG. 3 (b) shows an example of the equivalent circuit of the digital filter, and FIG. 3 (c) shows an example of the equivalent circuit of a plurality of digital filters. Indicates FIG. 4A is a graph showing an example of frequency characteristics of eight digital filters, and FIG. 4B is a graph showing an example of phase characteristics of eight digital filters. FIG. 5 (a) is a graph showing an example of the frequency characteristic of the digital filter, and FIG. 5 (b) is a graph showing an example of the phase characteristic of the digital filter. FIG. 6A is a schematic view showing an example of the configuration of the filter coefficient calculation device in the embodiment, and FIG. 6B is a schematic view showing an example of the function of the filter coefficient calculation device in the embodiment. FIG. 7 is a diagram showing an example of the flow of information processing of the filter coefficient calculation device in the embodiment. FIG. 8 is a diagram showing an example of part of the flow of information processing of the filter coefficient calculation device according to the embodiment. FIG. 9A is a graph showing an example of frequency characteristics of transfer function data, and FIG. 9B is a graph showing an example of phase characteristics of transfer function data. Fig.10 (a) is a graph which shows an example of the suppression amount of the digital filter which restrict | limited the upper limit, FIG.10 (b) is a graph which shows an 16-04-2019 10 example of the suppression amount of the digital filter which restrict | limited the lower limit. FIG. 11 (a) is a graph showing an example of the amount of suppression of the digital filter when the number of generations is 1, and FIG. 11 (b) is an example of the amount of suppression of the digital filter when the number of generations is 100. Is a graph showing FIG. 12 (a) is a graph showing an example of the amount of suppression of the digital filter when the number of generations is 300, and FIG. 12 (b) is an example of the amount of suppression of the digital filter when the number of generations is 500 Is a graph showing FIG. 13 (a) is a graph showing an example of the amount of suppression of the digital filter when the number of generations is 1000, and FIG. 13 (b) is an example of the amount of suppression of the digital filter when the number of generations is 10000 Is a graph showing FIG. 14 (a) shows a Nyquist diagram of the first reference data when the number of generations is 10000, and FIG. 14 (b) is a partially enlarged view of FIG. 14 (a). FIG. 15 is a schematic view showing an example of the configuration of the hearing aid in the embodiment. FIG. 16 is a schematic view showing an example of the configuration of a filter coefficient calculation device according to another embodiment. [0030] Hereinafter, an example of a filter coefficient calculation device and a hearing aid according to an embodiment of the present invention will be described with reference to the drawings. [0031] (Configuration of Filter Coefficient Calculation Device 1) An example of the configuration of the filter coefficient calculation device 1 in the present embodiment will be described with reference to FIG. FIG. 1 is a schematic view showing an example of the configuration of the filter coefficient calculation device 1 according to the present embodiment. [0032] As shown in FIG. 1, a personal computer (PC) is mainly used as the filter coefficient calculation device 1, and in addition, for example, an electronic device such as a portable terminal is used. In 16-04-2019 11 addition, the filter coefficient calculation device 1 may be incorporated in the negative feedback control device 5. The filter coefficient calculation device 1 is connected to the negative feedback control device 5 such as a hearing aid or an audio earphone, and is used to determine the digital filter coefficient of the negative feedback control device 5. [0033] The filter coefficient calculation device 1 may be connected to the negative feedback control device 5 via, for example, a public communication network (network). The public communication network indicates an Internet network or the like to which the filter coefficient calculation device 1 and the negative feedback control device 5 are connected via a communication circuit, and may be realized by an optical fiber communication network or a wireless communication network. [0034] The filter coefficient calculation device 1 calculates the transfer function E (s) using the signal input to the receiver 51 of the negative feedback control device 5 and the signal output from the microphone 52 in advance. The filter coefficient calculation device 1 determines the coefficients of the plurality of digital filters possessed by the control unit 53 connected between the receiver 51 and the microphone 52 based on the calculated transfer function E (s). For this reason, it is possible to suppress body noise and the like generated in the space 50 in the ear canal or the like by negative feedback control using a digital filter. [0035] The transfer function E (s) calculated by the filter coefficient calculation device 1 is expressed by the following [Equation 1]. Here, Y (s) represents a signal output from the microphone 52, and X (s) represents a signal input to the receiver 51. [0036] The transfer function E (s) can be shown, for example, as shown in FIG. FIG. 2 (a) is a graph showing the frequency characteristic of the transfer function E (s), and FIG. 2 (b) is a graph 16-04-2019 12 showing the phase characteristic of the transfer function E (s). The transfer function E (s) as shown in FIG. 2 has different characteristics depending on the frequency. For this reason, in order to realize optimal negative feedback control, it is necessary to use a plurality of digital filters. [0037] FIG. 3 is a diagram for explaining the principle of the negative feedback controller 5. FIG. 3 (a) shows an equivalent circuit of the negative feedback controller 5. FIG. 3 (b) shows an equivalent circuit of a digital filter. c) shows an equivalent circuit of a plurality of digital filters. [0038] As shown in FIG. 3A, the control unit 53 has a transfer function H (s) (a transfer function of the controller) set by a plurality of digital filters. That is, the filter coefficient calculation device 1 calculates the coefficients of a plurality of digital filters necessary for setting the transfer function H (s). [0039] The controller 53 receives the signal y (t). The control unit 53 converts the signal y (t) into a signal h (t) via a plurality of digital filters, and transmits the signal h (t) to the calculation unit 55. Arithmetic unit 55 receives signal r (t) as a reference signal from an external input (for example, a microphone) 54, and generates signal u (t) by dividing signal h (t) from signal r (t). . The negative feedback controller 5 receives the signal u ′ (t) converted through the transfer function E (s) and the signal d (t) resulting from external noise, noise generated in the ear, etc. A signal y (t) is generated by adding u ′ (t) and the signal d (t). Therefore, it is possible to output the signal y (t) in which the intensity of the signal d (t) caused by the noise or the like generated in the space 50 is suppressed. [0040] As shown in FIG. 3B, for example, a bi-quad filter F (bi-quadratic filter) is used as the digital filter. 16-04-2019 13 The bi-quad filter F includes five multiplication units FA1, FA2, FB0, FB1, and FB2, two delay units Z <−1>, and four calculation units FC1, FC2, FC3, and FC4. In addition, the number of each structure mentioned above is arbitrary, and you may set it as needed. [0041] As shown in FIG. 3C, a plurality of bi-quad filters F1, F2,..., FNb, and a multiplier GA are used as the plurality of digital filters of the controller 53. The biquad filters F1, F2, ..., FNb are connected in series. A multiplication unit GA is disposed between the bi-quad filter FNb and the operation unit 55. [0042] The filter coefficient calculation device 1 in the present embodiment can determine the values used for the multiplication units FA1, FA2, FB0, FB1, FB2 in the bi-quad filter F, and the multiplication unit GA. Details of this will be described later. [0043] FIG. 4 is a graph showing an example of frequency characteristics of eight digital filters, and FIG. 5 is a graph showing frequency characteristics when the eight digital filters shown in FIG. 4 are combined. FIGS. 4A and 5A show frequency characteristics of the digital filter, and FIGS. 4B and 5B show phase characteristics of the digital filter. [0044] In order to implement a digital filter having frequency characteristics as shown in FIG. 5, a plurality of digital filters as shown in FIG. 4 may be designed. In this case, it is necessary to determine the coefficients for each digital filter based on a very complicated relationship. Even in such a case, the filter coefficient calculation device 1 in the present embodiment can easily determine the optimal coefficient of the digital filter by using the genetic algorithm. 16-04-2019 14 [0045] In particular, it is desirable for the filter coefficient calculation device 1 in the present embodiment to use a real-valued genetic algorithm. By using the real-valued genetic algorithm, the inventors are able to determine the optimal coefficients of a plurality of digital filters in a short time, as compared to the case of using a genetic algorithm using ordinary bit operations. Found out. [0046] FIG. 6A is a schematic view showing an example of the configuration of the filter coefficient calculation device 1 according to the present embodiment. The filter coefficient calculation device 1 includes a housing 10, a CPU 101, a ROM 102, a RAM 103, a storage unit 104, and I / Fs 105 to 107. Each configuration 101-107 is connected by an internal bus 110. [0047] A CPU (Central Processing Unit) 101 controls the entire filter coefficient calculation device 1. A ROM (Read Only Memory) 102 stores an operation code of the CPU 101. A random access memory (RAM) 103 is a work area used when the CPU 101 operates. The storage unit 104 stores various types of information such as initial information. For example, a data storage device such as a solid state drive (SSD) other than a hard disk drive (HDD) is used as the storage unit 104. For example, the filter coefficient calculation device 1 may have a GPU (Graphics Processing Unit) not shown. By having a GPU, higher-speed arithmetic processing can be performed than usual. [0048] The I / F 105 is an interface for transmitting and receiving various information to and from an external device such as the negative feedback control device 5. The I / F 106 is an interface for transmitting and receiving information with the input unit 108. For example, a keyboard is used as the input unit 108, and a manager or operator of the filter coefficient calculation apparatus 1 inputs various information or control commands of the filter coefficient calculation apparatus 1 via the input unit 108. The I / F 107 is an interface for transmitting and receiving various information to and from the output unit 109. The output unit 109 outputs various information 16-04-2019 15 stored in the storage unit 104, the processing result of the filter coefficient calculation apparatus 1, and the like. A display is used as the output portion 109, and may include the input portion 108 as a touch panel, for example. [0049] FIG. 6B is a schematic view showing an example of the function of the filter coefficient calculation device 1 in the present embodiment. The filter coefficient calculation device 1 includes an acquisition unit 11, a population generation unit 12, an optimization unit 13, an output unit 14, an input unit 15, and an information DB 16. In the acquisition unit 11, the population generation unit 12, the optimization unit 13, the output unit 14, the input unit 15, and the information DB 16, the CPU 101 executes a program stored in the storage unit 104 or the like using the RAM 103 as a work area. Is realized by [0050] <Acquisition Unit 11> The acquisition unit 11 acquires the transfer function E (s). The acquisition unit 11 acquires transfer function data E ′ (jω) based on the transfer function E (s). The transfer function data E ′ (jω) is data obtained by performing correction processing on the transfer function E (s) as necessary. For example, the transfer function E (s) may be used as the transfer function data E ′ (jω). The details of the case of executing the correction process will be described later. [0051] <Population Generation Unit 12> The population generation unit 12 generates a population including a plurality of reference data. The plurality of reference data are set based on initial information acquired in advance. [0052] The reference data comprises data corresponding to the coefficients of the plurality of digital filters. For example, based on the center frequency, the quality factor, and the gain value of the 16-04-2019 16 reference data, the values of the multiplication units FA1, FA2, FB0, FB1, and FB2 illustrated in FIG. 2B and FIG. It can be set according to the number Nb of. The quality factor (quality factor) is generally called a Q value, and is a value representing the sharpness of the resonance peak. [0053] The initial information has a value for setting reference data in an initial stage, and has, for example, at least one of a central frequency distribution fc, a quality factor distribution Q, and a gain value distribution G. In this case, the center frequency, the quality factor, and the gain value are set, for example, based on the range of each distribution included in the initial information. [0054] The initial information includes, for example, the number Nb of digital filters and the number Np of coefficients of each digital filter, and also has initial conditions required when using a genetic algorithm, and values required for evaluation described later, etc. . [0055] <Optimization Unit 13> The optimization unit 13 uses a genetic algorithm to randomly select a part of reference data in the population, and updates the selected part of the reference data. The optimization unit 13 repeats the above selection and updating a plurality of times. The optimization unit 13 includes a child data generation unit 13a, an evaluation unit 13b, and an update unit 13c. [0056] <Child Data Generation Unit 13a> The child data generation unit 13a generates a plurality of child data ti based on the selected partial reference data (crossover). The child data generation unit 13a generates a plurality of child data ti based on, for example, REX (Real-coded Ensemble Xover). 16-04-2019 17 [0057] The optimization unit 13 generates a larger number of child data ti than the number of selected partial reference data. For this reason, when generating a plurality of child data ti, the possibility of becoming better data as coefficients of the digital filter can be increased as compared to some reference data. [0058] <Evaluation Unit 13b> The evaluation unit 13b generates evaluation data R (ω) having transfer function data E ′ (jω) and child data ti, and evaluates the evaluation data R (ω). The evaluation unit 13b repeats the above generation and evaluation. The evaluation data R (ω) has transfer function data E ′ (jω) and indicates a value corresponding to the amount of suppression of the digital filter. For this reason, the evaluation unit 13b can perform the evaluation based on the suppression amount corresponding to the environment in which the digital filter is used. [0059] <Update Unit 13c> The update unit 13c selects some child data ti among the plurality of child data ti based on the evaluation result in the evaluation unit 13b. The updating unit 13 c updates the selected partial child data ti as a partial reference data selected when generating the child data ti. The number to select and update the child data ti can be set arbitrarily, and may be selected and updated, for example, in the descending order of the evaluation result of the evaluation unit 13b. This allows the reference data in the population to be close to coefficients suitable for digital filters. [0060] <Output Unit 14> The output unit 14 transmits (inputs) the signal X (s) to the receiver 51 via the I / F 105. The output unit 14 selects first reference data from the population, and outputs coefficients of a plurality of digital filters based on the first reference data. The output unit 14 performs, for example, the same evaluation as the evaluation unit 13 b on a plurality of reference data, and evaluates the highest evaluation reference data as the first reference data. For this reason, it is possible to select the first reference data as the optimal digital filter coefficient from 16-04-2019 18 the population having the reference data updated by the optimization unit 13. [0061] The output unit 14 causes the output unit 109 to display the digital filter coefficients via the I / F 107. For example, the output unit 14 transmits the digital filter coefficients to the control unit 53 of the negative feedback controller 5 via the I / F 105. The filter coefficients may be set directly. [0062] <Input Unit 15> The input unit 15 acquires the signal Y (s) output from the microphone 52 via the I / F 105. The input unit 15 acquires, for example, initial information and the like input from the input unit 108 via the I / F 106. The input unit 15 may acquire initial information and the like via a storage medium such as, for example, a portable memory. [0063] <Information DB 16> The information DB 16 stores initial information acquired in advance, the signal X (s), the signal Y (s), the transfer function E (s), transfer function data E ′ (jω), and the like. In addition to the above, the information DB 16 stores a population including reference data generated later, child data ti, evaluation data R (ω), determination conditions, and the like. Various information (data) is stored in the storage unit 104. Each structure 11-15 stores various information in information DB16 as needed, or takes out various information. [0064] (Process of Filter Coefficient Calculation Device 1) Next, a procedure of information processing of the filter coefficient calculation device 1 in the present embodiment will be described. FIG. 7 is a diagram showing an example of the flow of information processing of the filter coefficient calculation device 1 in the present embodiment. 16-04-2019 19 [0065] <Acquisition of Transfer Function Data E ′ (jω): Step S110> As shown in FIG. 7, the acquiring unit 11 acquires transfer function data E ′ (jω) subjected to correction processing based on the transfer function E (s) ( Step S110). The acquisition unit 11 acquires, for example, the signal X (s) and the signal Y (s) in advance to calculate the transfer function E (s), and also acquires, for example, the transfer function E (s) calculated in advance. Good. [0066] The acquisition unit 11 stores, for example, the acquired transfer function E (s) and transfer function data E ′ (jω) in the information DB 16. The signal X (s) may be transmitted from the filter coefficient calculation device 1 to the negative feedback control device 5 or may be transmitted from the other terminal to the negative feedback control device 5, for example. [0067] The acquisition unit 11 may acquire transfer function data E ′ (jω) with reference to, for example, the following [Equation 2]. Here, j represents a complex number, ω represents an angular frequency, and s = jω. fl indicates the correction frequency (Hz) to be corrected, and β indicates the gradient of the phase. The initial information has a correction frequency fl and a slope β. [0068] By referring to the above [Equation 2], it is possible to control the amplitude characteristic and the phase characteristic in the transfer function data E ′ (jω) lower than the correction frequency fl. [0069] In particular, when a hearing aid is used as the negative feedback control device 5, the transfer function E (s) in the ear may change due to attachment and detachment of the hearing aid and opening and closing of the user's own mouth. 16-04-2019 20 This is because the gap between the hearing aid and the ear canal changes or the volume of the ear canal changes minutely. At this time, by setting the correction frequency fl and the gradient β, the change of the transfer function E (s) can be considered in advance, and the stability under actual use conditions can be further enhanced. [0070] For example, the suppression amount of the digital filter determined when the correction frequency fl is 20 Hz and the gradient β is 10 is shown in FIG. As shown in FIG. 9, at 20 Hz or less, the amplitude characteristics can be made flat (arrows in FIG. 9A), and the phase characteristics can be rotated (arrows in FIG. 9B). This makes it possible to stably drive the hearing aid even when the transfer function E (s) changes slightly due to the movement of the user's jaw or the like. [0071] <Generation of Population: Step S120> Next, the population generation unit 12 generates a population including a plurality of reference data set based on initial information acquired in advance (step S120). The population generation unit 12 acquires initial information from, for example, the information DB 16 and generates a population. The number of reference data is included in the initial information and can be set arbitrarily. The population generation unit 12 stores, for example, the acquired population in the information DB 16. [0072] The population generation unit 12 sets the number of coefficients of each reference data based on the number Nb of digital filters having initial information and the number Np of coefficients of each digital filter. The number Nn of coefficients of each reference data (number of coefficients to be identified) is represented by Nb × Np, and for example, when Nb = 8 and Np = 3, Nn = 24. [0073] 16-04-2019 21 The population generation unit 12 acquires, for example, a central frequency distribution fc represented by the following [Equation 3]. Here, ε ini indicates a random number having an arbitrary variance, and f cup indicates an upper limit frequency (Hz). The population generation unit 12 sets a center frequency based on the center frequency distribution fc and generates reference data. For this reason, the center frequency of each reference data can be set in a state where the setting frequency is weighted within the range indicated by the center frequency distribution fc. Thereby, when using a genetic algorithm, the search range can be limited in advance, and the updated reference data can be reliably converged. [0074] In particular, when a hearing aid is used as the negative feedback controller 5, the cutoff frequency of the digital filter can not be set to a negative value. Therefore, when the digital filter includes a low pass filter, setting the upper limit frequency fcup and the lower limit frequency makes it possible to converge to the center frequency applicable as the digital filter coefficient in the hearing aid. For example, by setting the upper limit frequency fcup to 8000 Hz and further setting the lower limit frequency to 1 Hz, it becomes easy to converge on the center frequency applicable as the coefficient of the digital filter in the hearing aid. [0075] The population generation unit 12 acquires, for example, a quality factor distribution Q indicated by the following [Equation 4] and a gain value distribution G indicated by the following [Equation 5]. Here, Qup indicates the upper limit quality factor, and Gup indicates the upper limit gain value (dB). The population generation unit 12 sets a quality factor based on the quality factor distribution Q, sets a gain value based on the gain value distribution G, and generates reference data. Therefore, the quality factor and the gain value can be limited, and the digital filter can be prevented from having a steep frequency characteristic. [0076] In particular, when a hearing aid is used as the negative feedback controller 5, a peaking filter, which is a type of digital filter, plays an important role in order to amplify or attenuate a desired band. The frequency characteristics of the peaking filter are determined by a center frequency 16-04-2019 22 that indicates the frequency of the peak, a quality factor that indicates the sharpness of the peak, and a gain value that controls the height of the peak. For this reason, when the digital filter includes a peaking filter, the formation of the extreme peak can be suppressed by setting the upper limit quality factor Qup and the upper limit gain value Gup. For example, by setting the upper limit quality factor Qup to 2 and the upper limit gain value Gup to 25, formation of unnecessary peaks in the hearing aid can be further suppressed. [0077] <Population optimization: Step S130> The optimization unit 13 selects a part of reference data in the population using a genetic algorithm, updates the selected part of reference data, selects and A plurality of updates are repeated (step S130). The optimization unit 13 carries out step S130 based on, for example, a JGG model (Just Generation Gap model). [0078] The optimization unit 13 acquires, for example, the population and the initial information from the information DB 16, and repeats selection and update of some reference data. The optimization unit 13 stores, for example, a population including updated reference data in the information DB 16, stores various data acquired in the process of updating as appropriate in the information DB 16, or takes out from the information DB 16 as appropriate. [0079] In step S130, steps S131 to S134 are repeated multiple times as selection and update. The optimization unit 13 may repeat steps S131 to S134 based on the number of repetitions (generation number Ngen) of the initial information, or may end the repetition based on, for example, comparison results of reference data before and after updating. . The optimization unit 13 appropriately extracts initial information required in steps S131 to S134 and various information generated from the information DB 16 or stores the information. [0080] 16-04-2019 23 <Selection of Partial Reference Data: Step S131> The optimization unit 13 randomly selects partial reference data from the population based on the number of parents μ included in the initial information (step S131). The parent number μ indicates the selection number of reference data, and is indicated by, for example, Nn + 1, and when Nn = 24, μ = 25. The optimization unit 13 obtains, for example, a matrix P of μ rows and Nn columns represented by the following [Equation 6], for example, as selected partial reference data. That is, the number of rows of matrix P indicates the number of selected reference data, and the number of columns of matrix P indicates the number of coefficients of each reference data. [0081] <Generation of a plurality of child data ti: Step S132> Next, the child data generation unit 13a generates a plurality of child data ti based on the selected partial reference data (Step S132). The child data generation unit 13a converts the matrix P into a distribution of random numbers N (0, 0.1) and calculates a matrix Pt represented by [Equation 8] below, for example, with reference to [Equation 7] below. Do. Here, Npоp indicates the group size of the initial information. For example, Npоp is represented by Nn × 25, and when Nn = 24, Npоp = 600. [0082] Next, the child data generation unit 13a sets a parent individual yi represented by the following [Equation 9] based on the matrix Pt. Here, i indicates one of the number of parents μ. [0083] Next, the child data generation unit 13a calculates the center of gravity <y> of the parent individual indicated by the following [Equation 10], and generates a child individual xi indicated by the following [Equation 11]. Here, ε indicates an arbitrary random number, and λ indicates the number of generator individuals. The initial information has a random number ε and a generator individual number λ. The genertor population λ is indicated by, for example, Nn × 4. [0084] 16-04-2019 24 Thereafter, the child data generation unit 13a converts the child individual xi with reference to the above [Equation 7] to generate child data ti shown in the following [Equation 12]. Here, fci, Qi, and Gi indicate the center frequency, the quality factor, and the gain value of the child data ti, respectively. The number of child data ti is indicated by λ, which is similar to the number of generator children. [0085] <Generation of Evaluation Data R (ω), Evaluation: Step S133> Next, the evaluation unit 13b generates evaluation data R (ω) including transfer function data E ′ (jω) and child data ti, and performs evaluation. The data R (ω) is evaluated, and a plurality of generation and evaluation are repeated (step S133). As in the case of using a conventional genetic algorithm, the evaluation unit 13b repeatedly evaluates the evaluation data R (ω) with an arbitrary evaluation function, and as shown in FIG. 8, for example, generates and evaluates step S133a. The step S133 f may be repeated. The evaluation unit 13b repeatedly performs the evaluation based on the number λ of the child data ti. Hereinafter, the case where steps S133a to S133f are performed will be described. [0086] <Generation of Transfer Function H (s) of Controller: Step S133a> The evaluation unit 13b generates a transfer function H (s) of the controller shown in the following [Equation 13] (Step S133a). The transfer function H (s) indicates the transfer function when the child data ti is used as a coefficient of the digital filter. The transfer function H (s) is separately generated for each child data ti. [0087] <Generation of open-loop transfer function L (s): Step S133 b> Next, the evaluation unit 13b generates an open-loop transfer function L (s) shown in the following [Equation 14] (Step S133 b). Here, Gall indicates the overall gain value (dB). An initial value of the entire gain value Gall is set in advance as initial information, and is, for example, -80 dB. The entire gain value Gall is updated according to the determination of each evaluation data R (ω) described later, and is then output as a coefficient of the multiplication unit GA shown in FIG. 3C. The overall gain value Gall 16-04-2019 25 is set to an initial value for each transfer function H (s). [0088] <Generation of Evaluation Data R (ω): Step S133c> Next, the evaluation unit 13b generates evaluation data R (ω) shown in the following [Equation 15] (Step S133c). The evaluation data R (ω) corresponds to the amount of suppression when the child data ti is used as a coefficient of the digital filter. The evaluation data R (ω) is generated for each open transfer function L (s). [0089] <Determining Evaluation Data R (ω): Step S133d> Next, the evaluation unit 13b includes the following [Equation 16] as a preset condition in the evaluation data R (ω) and the evaluation data R (ω). The loop transfer function L (s) is determined (step S133d). Here, fd1 represents the first frequency (Hz) serving as a reference for discrimination, plоw represents the upper limit (dB) below the first frequency fd1, and pup represents the upper limit (dB) above the first frequency fd1. . The initial information has a first frequency fd1 and upper limit values plow, pup. [0090] The evaluation unit 13b determines whether the evaluation data R (ω) satisfies the above condition, and when satisfied, updates the entire gain value Gall included in the evaluation data R (ω) (step S133e). For example, the evaluation unit 13b updates a value obtained by adding 1 to the entire gain value Gall as the entire gain value Gall, and executes Step S133b again. On the other hand, when the evaluation data R (ω) does not satisfy the above condition, the evaluation unit 13b executes the evaluation of the evaluation data R (ω) (step S133f). [0091] By the evaluation unit 13 b determining the evaluation data R (ω), it is possible to set the overall gain value Gall optimized for each of the evaluation data R (ω). Therefore, the optimum overall gain value Gall can be set for each child data ti generated by the genetic algorithm. Thereby, when using the determined coefficient of the digital filter, it is possible to suppress the signal 16-04-2019 26 which is matched with the phase of the original signal and the phase of the signal subjected to the negative feedback control and which becomes positive feedback, and an unintended signal It is possible to suppress the amplification of [0092] In particular, when a hearing aid is used as the negative feedback control device 5, amplification of positive feedback in the muffled sound suppression function can be controlled, and an increase in sound pressure can be suppressed. For example, the suppression amount of the digital filter determined in the case where the first frequency fd1 is 300 Hz, the upper limit ploww is 5 dB, and the upper limit pup is 5 dB is shown in FIG. As shown to Fig.10 (a), the amplitude of the whole suppression amount can be suppressed below a broken line (5 dB). [0093] The evaluation unit 13b may determine the evaluation data R (ω), for example, using the abovementioned [Equation 16] and the following [Equation 17] as preset conditions. Here, fd2 represents the second frequency (Hz) serving as a reference for determination, and nlow represents the lower limit (dB) of the second frequency fd2 or less. [0094] By determining the evaluation data R (ω) under the condition of the above [Equation 17], it is possible to arbitrarily set a frequency band in which the negative feedback control is unnecessary. [0095] In particular, when a hearing aid is used as the negative feedback control device 5, the battery consumption can be suppressed by setting the lower limit value nliw of the second frequency fd2 or less. For example, the suppression amount of the digital filter determined in the case where the second frequency fd2 is 10 Hz and the lower limit value nlow is −5 dB is shown in FIG. As 16-04-2019 27 shown in FIG. 10 (b), the amplitude of the suppression amount at 10 Hz or less can be suppressed to the broken line (-5 dB) or more. [0096] <Evaluation of Evaluation Data R (fi): Step S133f> Next, the evaluation unit 13b evaluates the evaluation data R (fi) with reference to the evaluation function E shown in the following [Equation 18] (Step S133f). ). Here, Rtg (fi) represents a target value (dB), C1 represents a first weighting factor of initial information, C2 represents a second weighting factor of initial information, and fi represents an evaluation frequency of initial information (Hz) is shown, and k is an arbitrary number. [0097] The initial information has a target value Rtg (fi), a first weighting factor C1, a second weighting factor C2, an evaluation frequency fi, and an arbitrary number k. By setting the weighting factors C1 and C2 shown in the above [Equation 18], it is possible to show an evaluation result which tends to converge on the target suppression amount. [0098] In particular, when a hearing aid is used as the negative feedback controller 5, for example, the first weight coefficient C1 is set to 0.1, and the second weight coefficient C2 is set to 100. In addition, the number k of the evaluation frequency fi is 5, the first evaluation frequency f1 is 100 Hz, the second evaluation frequency f2 is 140 Hz, the third evaluation frequency f3 is 200 Hz, and the fourth evaluation frequency f4 is 300 Hz. By setting the evaluation frequency f5 to 400 Hz, it is possible to show an evaluation result that easily converges to a suppression amount suitable for a hearing aid. [0099] <Updating a plurality of child data ti as a part of reference data: Step S134> After executing the repetition of generation and evaluation in step S133 described above (loop of child individual xi), 16-04-2019 28 the updating unit 13c performs the evaluation of step S133f. Based on the result, a part of child data ti is selected among the plurality of child data ti, and updated as a part of reference data selected in step S131 (step S134). The updating unit 13 c selects, for example, child data ti corresponding to the top μ evaluation data R (ω) in the evaluation result, and updates the child data ti as μ selection reference data. At this time, the updating unit 13 c updates, for example, the selected partial child data ti and the entire gain value Gall corresponding to the selected partial child data ti as the selected partial reference data. Or you may add. [0100] <Select First Data and Output Digital Filter Coefficient: Step S140> After executing repetition of selection and update (generation loop) in step S130 described above, the output unit 14 outputs first reference data from the population Are selected, and the coefficients of the plurality of digital filters are output based on the first reference data (step S140). The output unit 14 selects reference data having a coefficient closest to the target value Rtg (fi). [0101] The output unit 14 may evaluate the reference data of the population, for example, with reference to the evaluation function E shown in the above [Equation 18]. At this time, the output unit 14 may generate and evaluate the evaluation data R (ω) shown in the above [Equation 15], for example, based on each reference data. [0102] Thus, the procedure of the information processing of the filter coefficient calculation device 1 in the present embodiment ends. The initial information described above is stored in advance in the information DB 16 by the administrator, the worker, etc. via the input part 108 etc. For example, a database of initial information is constructed in advance in the storage unit 104, and information may be obtained as needed. You may retrieve from DB16. [0103] 16-04-2019 29 According to the filter coefficient calculation device 1 of the present embodiment, the evaluation unit 13b evaluates evaluation data R (ω) including transfer function data E ′ (jω) and child data ti using a genetic algorithm. . Further, the transfer function data E ′ (jω) is obtained based on the transfer function E (s) of the environment using the digital filter. For this reason, the evaluation unit 13b can perform the evaluation in consideration of the difference in the environment in which the digital filter is used. This makes it possible to easily determine the coefficients of the plurality of digital filters and realize optimization of the suppression amount in a short time. [0104] Further, according to the filter coefficient calculation device 1 in the present embodiment, the plurality of evaluation data R (ω) includes the entire gain value Gall, and the evaluation unit 13 b determines that the evaluation data R (ω) satisfies the condition. Updates the overall gain value Gall. That is, the optimal overall gain value Gall can be set for each child data ti generated by the genetic algorithm. Therefore, by using the entire gain value Gall as a coefficient of the digital filter, it is possible to adjust the entire suppression amount and to improve the accuracy of negative feedback control. [0105] Further, according to the filter coefficient calculation device 1 in the present embodiment, the evaluation unit 13 b generates the evaluation data R (ω) shown in the above [Equation 15]. At this time, the evaluation data R (ω) has the center frequency fci of the child data ti, the quality factor Qi, the gain value Gi, and the entire gain value Gall, and a plurality of digital filters using each value as a factor Corresponding to the amount of suppression of For this reason, the evaluation unit 13b can carry out an evaluation based on the actual suppression amount. This makes it possible to determine the coefficients of the plurality of digital filters with high accuracy. [0106] Further, according to the filter coefficient calculation device 1 in the present embodiment, the evaluation unit 13b determines the evaluation data R (ω) on the condition of the above [Equation 16], and based on the result of the determination, the evaluation data R (ω) Update or evaluate 16-04-2019 30 the The above [Equation 16] includes an expression for comparing the value of the evaluation data R (ω) with respect to the upper limit ploww equal to or lower than the first frequency fd1 and the upper limit pup equal to or higher than the first frequency fd1. For this reason, the upper limit value of the positive feedback in each evaluation data R (ω) (suppression amount) can be set arbitrarily. This makes it possible to suppress unintended signal amplification when the determined digital filter coefficients are used. [0107] Further, according to the filter coefficient calculation device 1 in the present embodiment, the evaluation unit 13b determines the evaluation data R (ω) on the condition of the above [Equation 16] and the above [Equation 17], and Based on the evaluation data R (ω) is updated or evaluated. The above [Equation 17] is an equation for comparing the value of the evaluation data R (ω) with respect to the lower limit value nlow of the second frequency fd2 or less. For this reason, when using a digital filter, it is possible to arbitrarily set a frequency band that does not require negative feedback control. This makes it possible to suppress the consumption of power. [0108] Further, according to the filter coefficient calculation device 1 in the present embodiment, the evaluation unit 13 b evaluates the evaluation data R (ω) with reference to the evaluation function E represented by the above [Equation 18]. The above [Equation 18] can indicate the difference between the target value Rtg (fi) at a plurality of evaluation frequencies fi and the evaluation data R (ω) as a unique value. Therefore, it is possible to easily select some child data ti among the plurality of child data ti based on the evaluation result, and to easily converge the child data ti to the target value Rtg (fi). This makes it possible to further shorten the time required to optimize the suppression amount. [0109] Further, according to the filter coefficient calculation device 1 in the present embodiment, the acquisition unit 11 acquires transfer function data E ′ (jω) based on the transfer function E (s) with reference to the above [Equation 2]. The above [Equation 2] is an expression that flattens the amplitude characteristic and rotates the phase characteristic in the transfer function E (s) below the correction frequency. For this reason, it is possible to determine the filter coefficient in 16-04-2019 31 consideration of the change of the transfer function E (s). This makes it possible to increase the stability of the suppression amount even when the transfer function E (s) in an environment using a digital filter changes. [0110] Further, according to the filter coefficient calculation device 1 in the present embodiment, the population generation unit 12 sets the center frequency based on the center frequency distribution fc. That is, the range of possible center frequencies of reference data can be limited in advance. For this reason, when using a genetic algorithm, the reference data to be updated can be reliably converged. This makes it possible to further shorten the time required to optimize the suppression amount. [0111] Further, according to the filter coefficient calculation device 1 in the present embodiment, the quality coefficient and the gain value are set based on the quality coefficient distribution Q and the gain value distribution G. Here, the frequency characteristics of the peaking filter included in the digital filter are determined by the frequency of the peak, the quality factor, and the gain value. Therefore, by limiting the upper limit of the quality factor and the gain value in advance, it is possible to eliminate the factor that produces the extreme peak. This makes it possible to improve the stability of the suppression amount when using a peaking filter as the digital filter. [0112] Next, an example of the filter coefficient calculation device 1 according to the present embodiment will be described. In the present embodiment, a hearing aid is used as the negative feedback controller 5. The hearing aid to be controlled has the transfer function E (s) (in-ear transfer function) shown in FIG. [0113] In the present embodiment, as the evaluation frequency fi, the first evaluation frequency f1 is 16-04-2019 32 100 Hz, the second evaluation frequency f2 is 200 Hz, the third evaluation frequency f3 is 300 Hz, and the fourth evaluation frequency f4 is 400 Hz. The target value Rtg (fi) is 20 dB at the first evaluation frequency f1, 20 dB at the second evaluation frequency f2, 17 dB at the third evaluation frequency f3, and 15 dB at the fourth evaluation frequency f4. [0114] In the present embodiment, eight bi-quad filters F are connected in series to the hearing aid. The eight bi-quad filters F used were five peaking filters, two low shelf filters, and one high pass filter. Further, each upper limit value plоw for suppressing amplification of positive feedback and pup are set to 5 dB. As the values of the other initial conditions, the values shown as an example above are appropriately used. [0115] The result of having implemented the filter coefficient calculation apparatus 1 by setting of the said initial information is shown in FIGS. 11-14. 11 to 13 show suppression amounts based on reference data updated at the end of different generation numbers Ngen. FIGS. 11 to 13 show 25 types of frequency characteristics, which correspond to the number of updated reference data. Also, the target value Rtg (fi) is indicated by a thick line. [0116] FIG. 11 (a) shows the results for the number of generations Ngen of 1, FIG. 11 (b) shows the results for the number of generations Ngen of 100, and FIG. 12 (a) shows the results for the number of generations Ngen of 300; (B) shows the result of the generation number Ngen of 500, FIG. 13 (a) shows the result of the generation number Ngen of 1000, and FIG. 13 (b) shows the result of the generation number Ngen of 10000. [0117] The results with generation numbers Ngen of 1 and 100 show that the difference in each frequency characteristic is noticeable as shown in FIG. 11, and also shows that each frequency characteristic deviates from the target value Rtg (fi). Show. 16-04-2019 33 On the other hand, when the generation number Ngen is 300, 500, and so on, as shown in FIG. 12 and FIG. 13A, the result that the difference between the respective frequency characteristics is small is shown and the target value Rtg (fi) Adjacent results are shown. Furthermore, as shown in FIG. 13B, when the number of generations Ngen is 10000, as shown in FIG. 13B, the results show that the frequency characteristics become almost equal, and the results that become almost equal to the target value Rtg (fi). [0118] Further, among the reference data when the generation number Ngen shown in FIG. 13 (b) is 10000, the reference data (first reference data) which takes the amount of suppression closest to the target value Rtg (fi) is selected. The results of determining the coefficients of the digital filter of the above correspond to the shapes shown in FIG. 4 and FIG. 5, and it was confirmed that very complex frequency characteristics were obtained. [0119] Further, the result of deriving the Nyquist diagram in the first reference data corresponds to the shape shown in FIG. Fig. 14 (a) shows a Nyquist diagram, and Fig. 14 (b) shows an enlarged view around (0, -1j) in Fig. 14 (a). Here, in FIG. 14, the Nyquist diagram in the first reference data is plotted outside the range of the broken line. The range enclosed by the dashed circle and straight line (hatched portion) indicates that the positive feedback is 5 dB or more. For this reason, it confirmed that positive feedback was 5 dB or less. [0120] (Configuration of Hearing Aid 100) Next, an example of the configuration of the hearing aid 100 in the embodiment will be described. FIG. 15 is a schematic view showing an example of the configuration of the hearing aid 100. As shown in FIG. The hearing aid 100 in the present embodiment includes the above-described filter coefficient calculation device 1 as a filter coefficient calculation unit. 16-04-2019 34 [0121] As shown in FIG. 15, the hearing aid 100 includes an external microphone 154, a hearing aid processor 155, an ear canal microphone 152, a receiver 151, and a controller 153. [0122] The external microphone 154 converts the sound transmitted from the external space into an electrical signal. The hearing aid processing unit 155 performs hearing aid processing such as gain adjustment according to the user. That is, the hearing aid processing unit 155 performs hearing aid processing on the electric signal converted by the external microphone 154 according to the user. The ear canal microphone 152 converts the sound transmitted from the inside of the ear canal 150 into an electrical signal. The receiver 151 outputs the sound converted from the electrical signal. The controller 153 includes a plurality of digital filters disposed between the receiver 151 and the ear canal microphone 152, and a filter coefficient calculation unit that determines the coefficients of the plurality of digital filters. The filter coefficient calculation unit has the same configuration as that of the filter coefficient calculation device 1 described above, and the detailed description will be omitted. [0123] According to the hearing aid 100 in this embodiment, the evaluation unit 13b uses the genetic algorithm to evaluate including the transfer function data E ′ (jω) and the child data ti, as in the filter coefficient calculation device 1 described above. Evaluate the data R (ω). Further, the transfer function data E ′ (jω) is obtained based on the transfer function E (s) of the environment using the digital filter. For this reason, the evaluation unit 13b can perform the evaluation in consideration of the difference in the environment in which the digital filter is used. This makes it possible to easily determine the coefficients of the plurality of digital filters and realize optimization of the suppression amount in a short time. [0124] (Configuration of Filter Coefficient Calculation Device 200) Next, an example of the configuration 16-04-2019 35 of the filter coefficient calculation device 200 in another embodiment will be described. FIG. 16 is a schematic view showing an example of the configuration of the filter coefficient calculation device 200. As shown in FIG. The difference between the filter coefficient calculation device 200 in the present embodiment and the above-described filter coefficient calculation device 1 is that it is used to determine the coefficients of digital filters related to negative feedback control other than sound. For this reason, description is abbreviate | omitted about the structure similar to the content mentioned above. [0125] As shown in FIG. 16, the filter coefficient calculation device 200 calculates the transfer function E (s) using the signal input to the transmission unit 251 and the signal output from the reception unit 252. The filter coefficient calculation device 200 determines the coefficients of a plurality of digital filters possessed by the negative feedback control unit 253 connected between the transmission unit 251 and the reception unit 252 based on the calculated transfer function E (s). Therefore, temperature control in the space 50, vibration control transmitted to the side surface of the space 50, and the like can be realized by negative feedback control using a digital filter. [0126] The filter coefficient calculation device 200 may be connected directly to the negative feedback control unit 253 or may be connected to the negative feedback control unit 253 via, for example, a public communication network. [0127] According to the filter coefficient calculation device 200 in the present embodiment, the evaluation unit 13b uses the genetic algorithm to obtain the transfer function data E ′ (jω), the child data ti, and the like, as in the filter coefficient calculation device 1 described above. Evaluate evaluation data R (ω) including Further, the transfer function data E ′ (jω) is obtained based on the transfer function E (s) of the environment using the digital filter. For this reason, the evaluation unit 13b can perform the evaluation in consideration of the difference in the environment in which the digital filter is used. This makes it possible to easily determine the coefficients of the plurality of digital filters and realize optimization of the suppression amount in a short time. 16-04-2019 36 [0128] While embodiments of the present invention have been described, the embodiments are presented by way of example only and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and the gist of the invention, and are included in the invention described in the claims and the equivalent scope thereof. [0129] 1, 200: filter coefficient calculation device 10: case 11: acquisition unit 12: population generation unit 13: optimization unit 13a: child data generation unit 13b: evaluation unit 13c: update unit 14: output unit 15: input unit 16 : Information DB 5: Negative feedback control device 50: Space 51: Receiver 52: Microphone 53: Control unit 54: External input 55, 56: Calculation unit 100: Hearing aid 101: CPU 102: ROM 103: RAM 104: Storage unit 105- 107: I / F 108: input portion 109: output portion 110: internal bus 151: receiver 152: microphone for external ear canal 153: controller 154: external microphone 155: hearing aid processor 251: transmitter 252: receiver 253: negative feedback control Part C1: first weighting factor C2: second weighting factor F1 to FN b: Bi-quad filter FA1, FA2: Multiplication part FC1 to FC4: Arithmetic part GA: Multiplication part Z <-1>: Delay part 16-04-2019 37

1/--страниц