JP2018197965A - Design support device of small speaker and design support method of speaker - Google Patents

Abstract

To provide a design support device capable of reducing time and cost required for design in the design of a small speaker.SOLUTION: An input device 10 is a device for allowing a user to input desired acoustic characteristic data about a small speaker. An arithmetic unit 20 uses a learned model for causing a computer to output a design parameter of the small speaker on the basis of the acoustic characteristic data, and calculates a design parameter from the acoustic characteristic data inputted by the input device 10. The learned model is composed of DNN learned so as to calculate a design parameter from the acoustic characteristic data by using a plurality of pieces of learning data each including the design parameter generated by computer simulation using an acoustic FDTD method and the acoustic characteristic data having the design parameter.SELECTED DRAWING: Figure 1

Description

  The present disclosure relates to a design support device for a small speaker and a design support method for a speaker, and more particularly to a technique for supporting design of a speaker used in a small device such as a smartphone.

  In recent years, the demand for small devices typified by smartphones has increased, and along with this, the demand for speakers built in small devices has also increased, and the size and complexity of the speakers have been increasing.

  Conventionally, a so-called acoustic equivalent circuit analysis method is known as a speaker design method. The acoustic equivalent circuit analysis method replaces mechanical vibration systems such as diaphragms and dampers, and acoustic vibration systems such as cabinets and horns with equivalent electrical circuits, and uses the knowledge of electrical circuits to target acoustic devices (speakers). Etc.). Such an acoustic equivalent circuit analysis method is described in Non-Patent Documents 1 and 2, for example.

Edited by Takeo Yamamoto, “Speaker System (above)”, Radio Technology Co., Ltd., July 15, 1977, p. 35-44 By Mitsuro Goro, "Acoustic Engineering (Electronics / Communication / Information Curriculum Series Oriented in the 21st Century)", Shosodo, March 1987, p. 32-39 Edited by Masahiro Toyoda, edited by the Acoustical Society of Japan, “The World of Sound Seen by the FDTD Method (Acoustic Science Series 14), Corona, December 16, 2015, p. 1-46

  In a small speaker that is used in a small device such as a smartphone, a diaphragm serving as a sound source is provided inside the casing, and the size and position of the sound hole provided in the casing, the diaphragm and the sound hole The size of the air chamber provided between the two is an important design parameter. However, since the acoustic equivalent circuit analysis method described above cannot take into account the size and position of the sound hole, the analysis method requires a large or medium-sized speaker that does not require special consideration of the size or position of the sound hole. However, it cannot be applied to the design of small speakers in which the size and position of the sound hole are important design parameters. For this reason, the design of a small speaker is performed by a cut-and-try design by an engineer who has sufficient experience, and reduction of the time and cost required for the design is an issue.

  The present disclosure has been made to solve such a problem, and an object thereof is to provide a design support apparatus capable of reducing the time and cost required for designing a small speaker.

  The design support device of the present disclosure is a design support device for a small speaker, and includes an input device and a calculation device. The input device is a device for a user to input acoustic characteristic data indicating the acoustic characteristics of a small speaker. The arithmetic unit uses the learned model for causing the computer to function so as to output the design parameters of the small speaker based on the acoustic characteristic data, and calculates the design parameters from the acoustic characteristic data input by the input device. Composed. Here, the trained model is generated by a computer simulation using an acoustic FDTD (Finite-Difference Time-Domain) method, each of which includes a plurality of design parameters and acoustic characteristic data of a small speaker having the design parameters. It is constituted by a deep neural network (hereinafter referred to as “DNN (Deep Neural Network)”) learned to calculate design parameters from acoustic characteristic data using learning data.

  Preferably, the small speaker includes a diaphragm serving as a sound source and a space in the housing formed inside the housing of the small speaker. The size of the divided grid in the computer simulation using the acoustic FDTD method is smaller in the order of the diaphragm, the space in the housing, and the acoustic propagation space outside the housing.

  Preferably, the design parameters included in each learning data are randomly generated within a predetermined range.

  Preferably, the acoustic characteristic data includes data indicating a frequency characteristic of sound pressure at a predetermined analysis point.

  Preferably, the acoustic characteristic data used as a plurality of learning data is selected according to a user's request, either data indicating the absolute frequency characteristic of the sound pressure or data indicating the frequency characteristic of the sound pressure level. The

  Preferably, the acoustic characteristic data further includes data indicating a frequency characteristic of the phase at the analysis point.

  Preferably, the design parameters include the size and location of sound holes formed in the small speaker and the size of the air chamber.

  Further, the design support method of the present disclosure is a speaker design support method, which includes generating a learned model for causing a computer to function so as to output speaker design parameters based on acoustic characteristic data; The method includes a step of inputting characteristic data by a user, and a step of calculating design parameters from the input acoustic characteristic data using a learned model. The step of generating a learned model includes a step of generating a plurality of learning data each including a design parameter and acoustic characteristic data of a speaker having the design parameter by computer simulation using an acoustic FDTD method, Generating a learned model by learning DNN so as to calculate design parameters from acoustic characteristic data using the learning data.

  In the design support device for the small speaker according to the present disclosure, the input is performed by the input device using the learned model for causing the computer to function so as to output the design parameters of the small speaker based on the acoustic characteristic data of the small speaker. Design parameters are calculated from the acoustic characteristic data. The learned model is composed of DNN learned to calculate design parameters from acoustic characteristic data. As a result, a small speaker can be designed without relying on a cut-and-try design by an engineer.

  Here, a plurality of learning data used for DNN learning, each including a design parameter and acoustic characteristics data of a small speaker having the design parameter, is generated by computer simulation using the acoustic FDTD method. As a result, in order to prepare a plurality of learning data, it is possible to prepare a large number of learning data used for DNN learning without actually creating various acoustic structures and performing actual measurement.

  As described above, according to the design support device for a small speaker of the present disclosure, various acoustic structures are actually created to prepare DNN learning data without relying on a cut-and-try design by an engineer. Therefore, the time and cost required for designing a small speaker can be greatly reduced.

It is a block diagram which shows the structure of the design assistance apparatus of the small speaker according to embodiment of this indication. It is sectional drawing of the small speaker designed using the design assistance apparatus shown in FIG. It is the top view which looked at the small speaker shown in FIG. 2 from the x direction. It is the figure which showed an example of the design parameter of the small speaker shown in FIG. 2, FIG. 3, and the range of each parameter. It is the figure which showed the frequency characteristic of the sound pressure (Pa) in the analysis point P0 shown in FIG. It is the figure which showed the frequency characteristic of the phase of the sound wave in the analysis point P0 shown in FIG. It is a whole block diagram of the learning system which produces | generates the learned model used in the arithmetic unit shown in FIG. It is a block diagram of the production | generation system which produces | generates the data for learning used for the learning system shown in FIG. It is an external view when a small speaker is seen from the x direction. It is an external view when a small speaker is seen from the x direction. It is a figure explaining an example of the idea of the grid setting by a condition setting part. It is a block diagram of DNN used for the learned model produced | generated by the DNN learning part. It is the figure which showed the input / output of DNN in this Embodiment. It is a flowchart explaining the production | generation procedure of the learned model used in the arithmetic unit shown in FIG. It is a flowchart explaining the procedure of the process performed in the arithmetic unit shown in FIG. It is the figure which showed the frequency characteristic of the absolute value (Pa) of a sound pressure. It is the figure which showed the frequency characteristic of a sound pressure level (dB).

  Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the drawings, the same or corresponding portions are denoted by the same reference numerals and description thereof will not be repeated.

<Overall configuration of design support device>
FIG. 1 is a block diagram illustrating a configuration of a design support apparatus for a small speaker according to an embodiment of the present disclosure. With reference to FIG. 1, this design support apparatus 1 designs a small speaker built in a small device typified by a smartphone using a computer. The design support apparatus 1 includes an input device 10, an arithmetic device 20, and a display device 30.

  The input device 10 is a device for a user of the design support apparatus 1 who is a designer of a small speaker to input data (acoustic characteristic data) indicating acoustic characteristics of the small speaker to be designed. The acoustic characteristics of a small speaker can be represented by the sound pressure at a predetermined analysis point, the frequency characteristics of the phase, or the like. In this embodiment, the user can input desired sound pressure frequency characteristics and phase frequency characteristics from the input device 10 for a small speaker to be designed. The acoustic characteristic data will be described in detail later.

  The arithmetic unit 20 includes a CPU (Central Processing Unit), a processing program and a ROM (Read Only Memory) that stores a learned model described later, a RAM (Random Access Memory) that temporarily stores data, and the like. . The learned model will be described in detail later, including its generation method. The learned model outputs the design parameters (sound hole dimensions, position, etc.) of the small speaker based on the acoustic characteristic data. It is a model for causing a computer (arithmetic unit 20) to function, and is configured by a DNN in which a causal relationship between acoustic characteristic data and design parameters is learned. The arithmetic unit 20 calculates design parameters of the small speaker from the acoustic characteristic data received from the input device 10 using the learned model.

  The display device 30 is a device for displaying the design parameters of the small speaker calculated by the arithmetic device 20 to the user. A user who has input desired acoustic characteristic data from the input device 10 can check on the display device 30 the design parameters calculated by the arithmetic device 20 or the structure of a small speaker having such design parameters. The display device 30 may be configured integrally with the input device 10 by a touch panel display or the like, for example.

<Description of design parameters and acoustic characteristics data>
First, the configuration and design parameters of a small speaker designed using the design support apparatus 1 shown in FIG. 1 and acoustic characteristic data indicating the acoustic characteristics of such a small speaker will be described.

  2 and 3 are configuration diagrams of a small speaker designed using the design support apparatus 1 shown in FIG. 2 is a cross-sectional view of the small speaker, and FIG. 3 is a plan view of the small speaker shown in FIG. 2 viewed from the x direction. The configuration of the small speaker shown in FIG. 2 and FIG. 3 is a model of the small speaker focused on important parameters (sound hole size, position, etc.) in the design of the small speaker.

  With reference to FIG. 2 and FIG. 3, the small speaker 200 designed using the design support apparatus 1 includes a housing 210, a diaphragm 220, a sound hole 230, and an air chamber 240. The acoustic propagation space 250 is a space through which sound waves output to the outside of the housing 210 through the sound hole 230 propagate. Note that the boundary 260 is an absorption boundary that is introduced in order to make the analysis target region finite in a computer simulation based on the acoustic FDTD method described later, and is, for example, a perfectly absorbing layer (PML).

  The diaphragm 220 is a sound source of the small speaker 200. That is, sound waves are generated when the vibration plate 220 vibrates. Although not particularly illustrated, a voice coil may be used as a driving source of the diaphragm 220 (electrodynamic conversion method), or a piezoelectric element that expands and contracts according to an applied voltage (piezoelectric conversion method). .

  The sound hole 230 is formed in the housing 210, and sound waves generated by the diaphragm 220 are output to the acoustic propagation space 250 outside the housing 210 through the sound hole 230. The air chamber 240 is provided between the space in which the diaphragm 220 is provided and the sound hole 230. In this embodiment, only one air chamber 240 is provided, but a plurality of air chambers may be provided in the x direction.

  LHx (FIG. 2) indicates the length of the sound hole 230 in the x direction. LHy (FIGS. 2 and 3) indicates the length of the sound hole 230 in the y direction, and LHz (FIG. 3) indicates the length of the sound hole 230 in the z direction. LCx (FIG. 2) indicates the length of the air chamber 240 in the x direction. Hymin and Hzmin indicate the position of the sound hole 230 in the casing 210. For example, if Hymin = 0 and Hzmin = 0, the position of the sound hole 230 is the lower left corner of the housing 210 in FIG.

  An analysis point P0 is set in the acoustic propagation space 250 (FIG. 2). The analysis point P0 is a point for measuring and analyzing the acoustic characteristics of the small speaker 200. The acoustic characteristics of the small speaker 200 are represented by the acoustic characteristics at the analysis point P0, and the desired acoustic characteristics are obtained at the analysis point P0. The small speaker 200 is designed so as to be obtained. In this embodiment, the analysis point P0 is set at a position away from the center of the sound hole 230 on the end surface of the casing 210 by L0 in the x direction.

FIG. 4 is a diagram showing an example of design parameters of the small speaker 200 shown in FIGS. 2 and 3 and ranges of each parameter. 2 and 3 together with FIG. 4, in this embodiment, the design parameters are LCx, LHx, LHy, LHz, Hymin, and Hzmin (all 10 ⁻³ m). That is, LCx indicates the size of the air chamber, LHx, LHy, and LHz indicate the size of the sound hole 230, and Hymin and Hzmin indicate the position of the sound hole 230. In the following description, these design parameters may be represented by a design variable vector X = [LCx, LHx, LHy, LHz, Hymin, Hzmin].

The range of each parameter indicates a designable range. In this embodiment, each range of LCx and LHx is 0.2 to 2.0 (10 ⁻³ m), and each range of LHy and LHz is 0.5 to 3.0 (10 ⁻³ m). Each range of Hymin and Hzmin is 0.5 to 11.5 (10 ⁻³ m).

  In the small speaker, the size and position of the sound hole with respect to the air chamber particularly affect the acoustic characteristics of the small speaker. Therefore, in this embodiment, the width of the air chamber 240 (the length in the y direction and the z direction). The design parameters are the width of the air chamber 240 in the x direction and the size and position of the sound hole 230.

  5 and 6 are diagrams showing examples of acoustic characteristic data of the small speaker 200. FIG. FIG. 5 shows the frequency characteristic of the sound pressure (Pa) at the analysis point P0 shown in FIG. 2, and FIG. 6 shows the frequency characteristic of the sound wave phase (rad) at the analysis point P0. Note that a Gaussian pulse is input to the diaphragm 220 (FIG. 2) (described later).

  With reference to FIG. 5 and FIG. 6, such a frequency characteristic profile (frequency waveform) of sound pressure and a phase frequency characteristic profile are examples of acoustic characteristic data indicating the acoustic characteristics of the small speaker 200. The frequency at which the sound pressure takes a peak (frequency at which the phase suddenly changes) indicates the resonance frequency of the small speaker 200, and such resonance frequency is also an example of acoustic characteristic data of the small speaker 200.

  Referring again to FIG. 1, the design support apparatus 1 is used for designing a small speaker 200 (FIGS. 2 and 3) indicated by the design parameters shown in FIG. In the design support apparatus 1, desired acoustic characteristic data (FIGS. 5 and 6) is input from the input device 10 by the user, and design parameters (FIG. 4) of the small speaker 200 indicating such acoustic characteristic data are calculated. Calculated by device 20.

  As described above, in the small speaker 200, the size and position (LHx, LHy, LHz, Hymin, Hzmin) of the sound hole 230, the size (LCx) of the air chamber 240, and the like are important design parameters. However, in the acoustic equivalent circuit analysis methods disclosed in Patent Documents 1 and 2 above, since the size and position of the sound hole cannot be taken into consideration, this analysis method takes special consideration of the size and position of the sound hole. Although it is useful for designing a large-sized or medium-sized speaker that is not necessary, it cannot be applied to the design of the small speaker 200 in which the size and position of the sound hole 230 are important design parameters.

  Therefore, the design support apparatus 1 according to this embodiment uses the learned model for causing the computer (the arithmetic unit 20) to function so as to output the design parameters of the small speaker 200 based on the acoustic characteristic data. The design parameters are calculated from the acoustic characteristic data input at 10. The learned model is composed of DNN learned to calculate design parameters from acoustic characteristic data. Thereby, the small speaker 200 can be designed without depending on the cut and try design by the engineer.

  Here, in order to accurately learn the causal relationship between acoustic characteristic data by DNN and design parameters, it is necessary to prepare a large number of learning data. Each learning data is composed of design parameters that are set as appropriate within the range that the design parameters can take (FIG. 4), and acoustic characteristic data of the small speaker 200 having the design parameters. In the design support apparatus 1 according to this embodiment, a plurality of learning data used for DNN learning is generated by computer simulation using the acoustic FDTD method. As a result, in order to prepare a large number of learning data, it is possible to prepare a large number of learning data used for DNN learning without actually creating various acoustic structures and performing actual measurements.

  Hereinafter, a learned model used in arithmetic device 20 of design support apparatus 1 according to this embodiment will be described.

<Description of trained model>
FIG. 7 is an overall block diagram of a learning system that generates a learned model used in the arithmetic device 20 shown in FIG. With reference to FIG. 7, the learning system 100 includes a database 120, a DNN learning unit 130, and a learning condition setting unit 140.

  The database 120 stores learning data 110-1, 110-2,..., 110 -N for generating a learned model in the DNN learning unit 130. Each of the learning data 110-1 to 110-N includes the above-described design parameter (design variable vector X) and acoustic characteristic data (sound pressure frequency characteristic and phase frequency characteristic) of the small speaker 200 having the design parameter. Paired data. The learning data 110-1 to 110-N have different design parameters within the range of each parameter shown in FIG.

  In order to generate a highly accurate learned model in the DNN learning unit 130, it is important to prepare a large number of learning data 110-1 to 110-N for various design parameters within a range that the design parameters can take. . In order to prepare a large number of learning data 110-1 to 110 -N, actually creating various acoustic structures and performing actual measurements have a great deal of time and cost. Therefore, in this embodiment, a large number of learning data 110-1 to 110-N are generated by computer simulation using the acoustic FDTD method.

  FIG. 8 is a configuration diagram of a generation system that generates learning data used in the learning system 100 shown in FIG. Referring to FIG. 8, generation system 300 includes a design parameter setting unit 310, an acoustic FDTD calculation unit 320, and a condition setting unit 330.

  The design parameter setting unit 310 generates a large number of different design parameters within the range shown in FIG. 4 in order to generate a large number of learning data 110-1 to 110 -N. For example, the design parameter setting unit 310 generates thousands of different design parameters.

  Generating a large number of learning data from a large number of design parameters can take a great deal of load and time. In this embodiment, the purpose is to reduce the calculation load of the acoustic FDTD calculation unit 320 that generates the learning data. In addition, the symmetry of the design parameters (line symmetry, point symmetry) is taken into consideration, thereby avoiding overlapping operations. Furthermore, the design parameter setting unit 310 randomly generates a large number of design parameters in order to increase the accuracy of a learned model generated using a large number of learning data.

  9 and 10 are external views when the small speaker 200 shown in FIGS. 2 and 3 is viewed from the x direction. In FIG. 9, the sound hole 230 is not shown. Referring to FIG. 9, it is assumed that the small speaker 200 has a square outer shape when the small speaker 200 is viewed from the x direction. In this case, considering the symmetry of the shape, it is sufficient to prepare the design parameters included in the region 270 in which the center of the sound hole 230 is indicated by hatching with respect to the position of the sound hole 230. There is no need to generate design parameters included in other areas.

  Therefore, the design parameter setting unit 310 generates a large number of design parameters for the design parameter in which the center of the sound hole 230 is included in the region 270 in consideration of the symmetry of the shape of the small speaker 200. Thus, by considering the symmetry of the design parameters, it is possible to reduce the calculation load and calculation time of the acoustic FDTD calculation unit 320 that calculates the acoustic characteristic data for each generated design parameter.

  Furthermore, referring to FIG. 10, design parameter setting section 310 performs random processing within the range that the design parameter can take (within the range where the center of sound hole 230 is included in region 270 in FIG. 9 in consideration of symmetry). A large number of design parameters are generated. “Random” means that the design parameter is arbitrarily changed instead of changing the design parameter at a predetermined step size. By using a large number of learning data based on randomly changed design parameters, it is possible to improve the accuracy of the learned model by avoiding learning falling into a local solution when generating a learned model by DNN. .

  Referring to FIG. 8 again, the acoustic FDTD calculation unit 320 includes a CPU, a GPU (Graphics Processing Unit), a ROM that stores a processing program, a RAM that temporarily stores data, and the like. The acoustic FDTD calculation unit 320 receives a large number of design parameters generated by the design parameter setting unit 310, and the small speaker 200 has the design parameters for each design parameter by computer simulation using the acoustic FDTD method. The acoustic characteristic data in the case is calculated.

  The FDTD method is a method that has been used for electromagnetic wave propagation analysis. The FDTD method differentiates Maxwell's equations, which are the dominant equations of electromagnetic waves, in time and space, and uses the Leapfrog algorithm to temporally change the electromagnetic field in the analysis space. This is a technique for obtaining the time response of the output point by updating. The acoustic FDTD method is an application of the FDTD method to sound wave propagation analysis, and wave propagation is calculated using sound pressure and particle velocity. This acoustic FDTD method can perform acoustic characteristic analysis in consideration of the shape and positional relationship of sound holes and air chambers. Hereinafter, an outline of the acoustic FDTD method will be described (for details, see Patent Document 3 above).

  In the acoustic FDTD method, the sound pressure and the particle velocity in the analysis space are spatially and temporally used by using a structural grid called staggered grids in which the reference points of the sound pressure and the particle velocity are alternately arranged spatially and temporally. Is discretized. Then, using the leapfrog algorithm, sound pressure and particle velocity discretized spatially and temporally are calculated alternately.

  In the acoustic FDTD method, the motion and deformation of air particles (lattice microvolume) in the analysis space are governed by the following equations.

  Equations (1) to (3) are equations of motion of air particles in the x, y, and z directions, respectively, and equation (4) is a continuous equation relating to sound pressure. ρ represents air density, and u, v, and w represent particle velocities in the x, y, and z directions, respectively. P represents sound pressure, and κ represents volume modulus. The above equations (1) to (4) are collectively referred to as “dominance equations”.

  By solving the above governing equation, the sound pressure and particle velocity in the analysis space can be obtained. In order to realize this on the computer, the acoustic FDTD method discretizes the above governing equation with respect to space and time. To do. Specifically, regarding the sound pressure, when the spatial discretization widths in the x, y, and z directions are Δx, Δy, and Δz and the temporal discretization width is Δt, respectively, regarding the particle velocity, Δx / 2, Δy / 2, Δz / 2, and time are discretized by being shifted from the sound pressure reference point by Δt / 2.

  When the above governing equations accompanying such discretization are approximated by difference, the governing equations of the equations (1) to (4) are respectively transformed as the following equations.

i, j, and k are spatial steps indicating the sound pressure reference points in the x, y, and z directions in the analysis space, respectively. n is a time step indicating the reference point with respect to time. For example, u ^{n + 1} (i + 1/2, j, k) represents the value of the particle velocity in the x direction when the spatial step is i + 1/2, j, k and the time step is n + 1, and p ^{n + 1 / 2} (i, j, k) indicates the sound pressure value when the spatial step is i, j, k and the time step is n + 1/2.

  By alternately calculating the particle velocity represented by the equations (5) to (7) and the sound pressure represented by the equation (8) with respect to the analysis space (all space steps), the states in the new time steps are successively calculated. (Leapfrog algorithm).

  Referring to FIG. 8, condition setting unit 330 sets calculation conditions for computer simulation using the acoustic FDTD method in acoustic FDTD calculation unit 320. Specifically, the condition setting unit 330 includes a sound pressure p, a bulk modulus κ, a spatial discretization width Δx, Δy, Δz, a time discretization width Δt, a diaphragm 220 that is a sound source (FIG. 2). ), The number of time steps, and other constants such as the speed of sound are set in the acoustic FDTD calculation unit 320. Note that the casing 210 (FIG. 2) is configured by a rigid body, and the sound wave is set to be totally reflected.

  In this embodiment, regarding the input to the diaphragm 220, a Gaussian pulse expressed by the following equation is given to the diaphragm 220, and acoustic characteristic data when the diaphragm 220 is excited by the Gaussian pulse is obtained. Calculated by the acoustic FDTD calculation unit 320.

g (t) = A · exp [− {(nΔt−T) /0.29T} ² ] (9)
T = 0.646 / fc (10)
In the above, A is the maximum amplitude value, and fc is the frequency at which the frequency spectrum of the amplitude of the Gaussian pulse decreases by 3 dB.

  The acoustic FDTD calculation unit 320 performs computer simulation using the acoustic FDTD method for each of a large number of randomly generated design parameters received from the design parameter setting unit 310 according to the calculation conditions set by the condition setting unit 330. Thus, the sound pressure and particle velocity in the analysis space (the space within the boundary 260 of the small speaker 200 shown in FIGS. 2 and 3) are sequentially calculated. The acoustic FDTD calculation unit 320 acquires time response data of sound pressure at the analysis point P0 shown in FIG. 2 for each design parameter, and performs fast Fourier transform (FFT) processing on the time response data. To calculate acoustic characteristic data (sound pressure frequency characteristic and phase frequency characteristic data at the analysis point P0).

  The acoustic characteristic data calculated in this way constitutes learning data in pairs with the design parameters corresponding to the acoustic characteristic data, and a large number corresponding to a large number of design parameters generated by the design parameter setting unit 310. The learning data 110-1 to 110-N are stored in the database 120 (FIG. 7).

  Note that the spatial discretization widths Δx, Δy, and Δz (hereinafter referred to as “grids”) set by the condition setting unit 330 are analysis spaces (FIGS. 2 and 3). In the illustrated space within the boundary 260 of the small speaker 200), it is preferable that the size (roughness) of the grid is appropriately changed according to the region.

  FIG. 11 is a diagram for explaining an example of the concept of grid setting by the condition setting unit 330. FIG. 11 is a cross-sectional view of the small speaker 200, in which a schematic grid is superimposed on the cross-sectional view of FIG.

  Referring to FIG. 11, condition setting unit 330 includes region 410 in the vicinity of diaphragm 220 as the sound source, region 420 excluding region 410 in case 210, and region 430 outside case 210 (acoustic propagation space 250). Change the grid size (roughness). Specifically, the grid size in the region 410 including the diaphragm 220 is relatively small, and the grid size in the region 430 (acoustic propagation space 250) outside the housing 210 is relatively large. A grid is set.

  As described above, the accuracy of the computer simulation by the acoustic FDTD method can be improved by changing the size of the grid appropriately according to the region, instead of making the size of the grid uniform in the analysis space. Can be shortened.

  Note that the three-stage grid division as shown in FIG. 11 is an example, and the grid division method is not limited to that shown in FIG. In addition, the example shown in FIG. 11 is schematically shown to explain the relative size difference between the three levels of grids, and does not indicate the absolute size of each grid. Absent.

  Referring to FIG. 7 again, DNN learning unit 130 includes a CPU, a ROM for storing processing programs, a RAM for temporarily storing data, and the like. Then, the DNN learning unit 130 performs DNN learning using a large number of learning data 110-1 to 110-N stored in the database 120, and calculates design parameters of the small speaker 200 from the acoustic characteristic data. A learned model for causing the computer (the arithmetic device 20 in FIG. 1) to function is generated. That is, this learned model is actually DNN in which the causal relationship between acoustic characteristic data and design parameters is learned using learning data 110-1 to 110 -N stored in database 120.

  FIG. 12 is a configuration diagram of the DNN used for the learned model generated by the DNN learning unit 130. Referring to FIG. 12, DNN includes an input layer, an output layer, and a plurality of intermediate layers. The DNN has a plurality of intermediate layers and thus has a high generalization performance compared to the conventional neural network (NN), and is particularly useful for a model having a complicated shape (design parameter) such as the small speaker 200. It is.

  The data input to the input layer is weighted at the joint portion with the first intermediate layer and input to the first intermediate layer. Each unit (neuron) of the first intermediate layer is provided with an activation function, and the output of the activation function of each unit is connected to the next second intermediate layer (not shown). Weighted and input to the second intermediate layer. Such data propagation is sequentially performed in a plurality of intermediate layers, and finally data is output to the output layer.

  The output data of the output layer is compared with the teacher data. Then, the gradient to the input layer is calculated based on the error between the output data of the output layer and the teacher data, and the weights (parameters) between the units are updated using the calculated gradient (error back propagation method). . By performing such processing with a large number of learning data, the causal relationship between input and output is modeled (generation of a learned model).

  FIG. 13 is a diagram showing the input / output of DNN in the present embodiment. Referring to FIG. 13, acoustic characteristic data of learning data received from database 120 (FIG. 7) is input to the input layer of DNN. In this embodiment, a frequency characteristic profile of an absolute value (Pa) of a sound pressure and a profile of a frequency characteristic of a phase are used as acoustic characteristic data. As an example, each profile of the frequency characteristics of sound pressure and phase is discretized with a predetermined frequency width Δf and applied to the input layer of the DNN.

  Data corresponding to the design parameter is output to the output layer. That is, the number of units in the output layer is the same as the number of variables in the design variable vector X = [LCx, LHx, LHy, LHz, Hymin, Hzmin] (six in this embodiment). The output data of the output layer is compared with design parameters (teacher data) paired with the acoustic characteristic data of the learning data input to the input layer, and weights (parameters) between the units based on the comparison error Is updated. The DNN learning unit 130 generates a learned model by performing the above-described processing on a large number of learning data stored in the database 120 (FIG. 7).

  Referring to FIG. 7 again, learning condition setting unit 140 sets learning conditions for generating a learned model in DNN learning unit 130. For example, the learning condition setting unit 140 sets the number of DNN intermediate layers (in this embodiment, four layers), and the number of units (number of neurons) in each of the input layer, each intermediate layer, and the output layer. . Further, for example, the learning condition setting unit 140 sets a condition for ending learning with each learning data. Specifically, test data (design parameters and acoustic characteristic data) different from the learning data is prepared in order to prevent over-learning of DNN by the learning data. In the learning using each learning data, test data (acoustic characteristic data) is given to the input layer of the DNN every time learning is performed, and an error (test error) between the output data of the output layer and the test data (design parameter). If the minimum value of the learning data is not continuously updated a predetermined number of times, the above predetermined number of times is set as a condition for ending the learning with each learning data so that the learning of the learning data ends (early end).

  FIG. 14 is a flowchart for describing a learning model generation procedure used in the arithmetic unit 20 shown in FIG. Referring to FIGS. 7 and 8 together with FIG. 14, first, a large number of learning data 110-1 to 110-N are generated in the processing from step S10 to S40. That is, in the generation system 300, design parameters for generating learning data are set by the design parameter setting unit 310 (step S10). The design parameter setting unit 310 considers the symmetry of the design parameter as described with reference to FIG. 9, and sets the design parameter randomly as described with reference to FIG.

  Next, the acoustic FDTD calculation unit 320 calculates acoustic characteristic data corresponding to the design parameter set in step S10 by computer simulation using the acoustic FDTD method (step S20). Then, learning data that is a pair of the design parameter set in step S10 and the acoustic characteristic data calculated in step S20 based on the design parameter is stored in the database 120 of the learning system 100 (step S30). .

  Next, the acoustic FDTD calculation unit 320 determines whether or not the number of learning data stored in the database 120 is equal to or greater than a predetermined number N (step S40). The predetermined number N is a value of several thousand levels, for example. If it is determined in step S40 that the number of learning data is less than the predetermined number N (NO in step S40), the process returns to step S10, considering the symmetry of the design parameters, and randomly Design parameters are set.

  If it is determined in step S40 that the number of learning data is greater than or equal to the predetermined number N (YES in step S40), then a learned model is generated in the processing from step S50 to S70. That is, in the learning system 100, the learning condition setting unit 140 sets DNN learning conditions in the DNN learning unit 130 (step S50). As described above, the learning conditions include, for example, the DNN configuration (number of layers, number of units (number of neurons), etc.), conditions for ending learning with each learning data, and the like.

  Next, in the DNN learning unit 130, the causal relationship between the acoustic characteristic data and the design parameters is learned by the DNN using the large number of learning data 110-1 to 110-N stored in the database 120 (steps). S60). Specifically, acoustic characteristic data of learning data is given to the input layer of the DNN, and a large number of learning data 110-1 to 110 are set using the design parameters of the learning data paired with the acoustic characteristic data as teacher data. DNN is learned using -N.

  When DNN learning using the learning data 110-1 to 110-N is completed, a learned model (actually a parameter indicating the DNN structure and weights between the units) is output from the DNN learning unit 130. (Step S70).

  FIG. 15 is a flowchart for explaining the procedure of processing executed in the arithmetic unit 20 shown in FIG. Referring to FIG. 15, the arithmetic unit 20 receives acoustic characteristic data (sound pressure (Pa) and phase frequency characteristic profiles as shown in FIGS. 5 and 6) input by the user from the input device 10. Obtained from the input device 10 (step S110). The acoustic characteristic data is not the acoustic characteristic data of the learning data stored in the database 120 of the learning system 100 but desired acoustic characteristic data input from the input device 10 by a user who will design a small speaker. .

  Next, the arithmetic unit 20 inputs the acoustic characteristic data acquired in step S10 to a learned model (post-learning DNN) generated in advance according to the procedure shown in FIG. 14, and uses the learned model to design parameters ( LCx, LHx, LHy, LHz, Hymin, Hzmin shown in FIGS. 2 and 3 are calculated (step S120).

  Then, the arithmetic unit 20 outputs the calculated design parameters LCx, LHx, LHy, LHz, Hymin, Hzmin to the display device 30 (step S130). In the display device 30, the numerical values of the design parameters received from the arithmetic device 20 may be displayed as they are, or the structure of the small speaker 200 having such design parameters may be illustrated.

  As described above, in this embodiment, an input using a learned model for causing a computer (the arithmetic unit 20) to function so as to output design parameters of a small speaker based on acoustic characteristic data of the small speaker. Design parameters are calculated from the desired acoustic characteristic data input by the apparatus 10. The learned model is composed of DNN learned to calculate design parameters from acoustic characteristic data. As a result, a small speaker can be designed without relying on a cut-and-try design by an engineer.

  A plurality of learning data 110-1 to 110-N used for DNN learning, each including a design parameter and acoustic characteristic data of a small speaker having the design parameter, are computer simulations using an acoustic FDTD method. Generated by. Thus, in order to prepare a plurality of learning data, a large number of learning data 110-1 to 110 -N used for learning the DNN are prepared without actually creating various acoustic structures and performing actual measurement. be able to.

  As described above, according to the design support device 1 for the small speaker of the present disclosure, various acoustic structures are actually created in order to prepare DNN learning data without relying on a cut-and-try design by an engineer. Thus, since it is not necessary to perform actual measurement, the time and cost required for designing a small speaker can be greatly reduced.

  In this embodiment, the size (roughness) of the grid in the computer simulation using the acoustic FDTD method is appropriately changed depending on the region. Specifically, the grid in the area near the diaphragm 220 as the sound source is reduced, and the grid in the acoustic propagation space 250 outside the housing is increased (FIG. 11). As a result, the accuracy of the acoustic FDTD method can be increased and the calculation time can be shortened as compared with the case where the size of the grid is uniform.

  Further, according to this embodiment, the analysis space is limited in consideration of the symmetry of the shape of the design target (small speaker 200), so that the acoustic FDTD computing unit 320 that calculates acoustic characteristic data for each design parameter. Calculation load and calculation time can be reduced.

  Furthermore, according to this embodiment, since DNN learning is performed using a large number of learning data based on design parameters that are randomly changed, learning becomes a local solution when a learned model is generated by DNN. The accuracy of the learned model can be improved by avoiding falling.

  In the above embodiment, the learning data 110-1 to 110-N used for generating the learned model is the acoustic characteristic data given to the input layer of the DNN during learning by the DNN learning unit 130. The sound pressure frequency characteristic is a frequency characteristic profile of the absolute value (Pa) of the sound pressure as shown in FIG. 16, but instead of the profile of the frequency characteristic of the absolute value (Pa) of the sound pressure, FIG. A frequency characteristic profile of the sound pressure level (dB) as shown may be applied to the DNN input layer. The frequency characteristic profile of the sound pressure level (dB) shows characteristics at frequencies other than the resonance frequency fr better than the frequency characteristic profile of the absolute value (Pa) of the sound pressure, and emphasizes the frequency characteristic profile. When designing, the accuracy of the learned model can be increased.

  In addition, when generating a learned model, the frequency characteristic profile of the sound pressure absolute value (Pa) is applied to the DNN input layer for learning, or the frequency characteristic profile of the sound pressure level (dB) is DNN. The user may be able to select whether to learn by giving to the input layer. For example, referring to FIG. 7 again, information indicating the user's selection result is set from the learning condition setting unit 140 to the DNN learning unit 130, and learning is performed using the frequency characteristic profile of the sound pressure level (dB). Is selected by the user, the frequency characteristic data of the absolute value (Pa) of the sound pressure received from the database 120 may be converted into decibels (dB) and applied to the DNN input layer.

  When a learned model is generated by applying a frequency characteristic profile of the sound pressure level (dB) to the input layer of the DNN, desired acoustic characteristics to be given to the learned model in the arithmetic unit 20 of the design support apparatus 1 The sound pressure frequency characteristic in the data also needs to be a profile of the sound pressure level (dB) frequency characteristic.

  In the above embodiment, the acoustic characteristic data is composed of data indicating the frequency characteristic of the sound pressure and data indicating the frequency characteristic of the phase, but the data indicating the frequency characteristic of the sound pressure. It is good only as well. As in the above-described embodiment, the combined use of not only the sound pressure frequency characteristic but also the phase frequency characteristic is expected to improve the DNN learning accuracy and the design accuracy of a small speaker using a learned model. it can.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present disclosure is shown not by the above description of the embodiments but by the scope of claims, and is intended to include all modifications within the meaning and scope equivalent to the scope of claims.

  DESCRIPTION OF SYMBOLS 1 Design support apparatus, 10 input apparatus, 20 arithmetic unit, 30 display apparatus, 100 learning system, 110-1, 110-2, ..., 110-N learning data, 120 database, 130 DNN learning part, 140 learning Condition setting unit, 200 small speaker, 210 housing, 220 diaphragm, 230 sound hole, 240 air chamber, 250 acoustic propagation space, 260 boundary, 270, 410 to 430 region, 300 generation system, 310 design parameter setting unit, 320 Acoustic FDTD calculation unit, 330 condition setting unit.

Claims (8)

A design support device for a small speaker,
An input device for a user to input acoustic characteristic data indicating the acoustic characteristics of the small speaker;
The design parameter is calculated from the acoustic characteristic data input by the input device using a learned model for causing a computer to function to output the design parameter of the small speaker based on the acoustic characteristic data. With a configured arithmetic unit,
The learned model is generated by computer simulation using an acoustic FDTD method, and a plurality of learning data each including the design parameter and the acoustic characteristic data of a small speaker having the design parameter, A small speaker design support apparatus configured by a deep neural network learned to calculate the design parameters from acoustic characteristic data. The small speaker is configured to include a diaphragm serving as a sound source, and a housing internal space formed inside the housing of the small speaker,
2. The small speaker design support according to claim 1, wherein the size of the divided grid in the computer simulation using the acoustic FDTD method is smaller in the order of the diaphragm, the space inside the housing, and the acoustic propagation space outside the housing. apparatus.   The design support device for a small speaker according to claim 1 or 2, wherein the design parameter included in each of the plurality of learning data is randomly generated within a predetermined range.   The design support apparatus for a small speaker according to any one of claims 1 to 3, wherein the acoustic characteristic data includes data indicating a frequency characteristic of sound pressure at a predetermined analysis point.   As the acoustic characteristic data used as the plurality of learning data, either data indicating the frequency characteristic of the absolute value of the sound pressure or data indicating the frequency characteristic of the sound pressure level is selected according to a user's request. 5. A small speaker design support apparatus according to claim 4.   The small-sized speaker design support apparatus according to claim 4, wherein the acoustic characteristic data further includes data indicating a frequency characteristic of a phase at the analysis point.   The design support device for a small speaker according to any one of claims 1 to 6, wherein the design parameters include a size and a position of a sound hole formed in the small speaker and a size of an air chamber. A speaker design support method,
Generating a learned model for causing a computer to function to output design parameters of the speaker based on acoustic characteristic data indicative of acoustic characteristics of the speaker;
A user inputting the acoustic characteristic data;
Using the learned model to calculate the design parameters from the inputted acoustic characteristic data,
Generating the trained model comprises:
Generating a plurality of learning data each including the design parameter and the acoustic characteristic data of a speaker having the design parameter by computer simulation using an acoustic FDTD method;
Generating a learned model by learning a deep neural network so as to calculate the design parameter from the acoustic characteristic data using the plurality of learning data, and a speaker design support method.

Images