JP7622852B2 - Estimation device, estimation method, and program - Google Patents

Description

The present invention relates to an estimation technique for estimating the reflectance of walls and other objects.

There are many devices that use point cloud sensors, such as LiDAR (Light Detection and Ranging or Laser Imaging Detection and Ranging). For example, Non-Patent Document 1 describes a device that can acquire point cloud data inside and outside a building. Here, point cloud data refers to data that represents the presence or absence of a point for each coordinate in a specific real-world space as a three-dimensional coordinate space. Point cloud data can also include additional information in addition to the presence or absence of a point for each coordinate, but the additional information is generally information related to the color of the point.

"What is 3D laser scanning measurement?", [online], Besterra Co., Ltd., [Retrieved July 21, 2021], Internet <URL: https://besterra3d.com/measure/about.html>

When trying to simulate an acoustic signal in a given space using point cloud data, it is necessary to consider not only the presence or absence of points but also the reflectivity of the points. Reflectivity represents the attenuation of the wave of a waveform when sound is reflected. For example, when sound hits a wall with a reflectivity of 0.5, a wave with half the amplitude of the wave before reflection is reflected from the wall. When trying to measure such reflectivity in real space, impulse response is sometimes used, but this requires a skilled person to spend time on it.

The present invention aims to provide a technology that utilizes simulation space and real space to enable estimation of reflectance in a short period of time.

In order to solve the above problem, according to one aspect of the present invention, an estimation device estimates the reflectance of a wall surface in a predetermined space. The estimation device includes a simulation acquisition unit that acquires a value based on the reverberation of an acoustic signal emitted from a predetermined position in a simulation space that simulates the predetermined space, a real space acquisition unit that acquires a value based on the reverberation of an acoustic signal emitted from a position corresponding to the predetermined position in the predetermined space, which is a real space, and a parameter estimation unit that estimates the reflectance of the wall surface using an evaluation value that uses the difference between the value based on the reverberation in the simulation space and the value based on the reverberation in the real space.

The present invention has the effect of enabling reflectance to be estimated in a short period of time.

FIG. 1 is a functional block diagram of an estimation system according to a first embodiment. FIG. 2 is a diagram showing an example of a processing flow of the estimation system according to the first embodiment. FIG. FIG. 11 is a diagram showing an example of a processing flow of an acquisition unit. FIG. 11 is a diagram illustrating an example of point cloud data after noise has been removed. FIG. 11 is a diagram illustrating an example of point cloud data after noise has been removed. FIG. 4 is a functional block diagram of a room shape creation unit. FIG. 11 is a diagram showing an example of a processing flow of a room shape creation unit. FIG. 11 is a diagram for explaining the processing of a loss completion unit. FIG. 11 is a diagram for explaining the processing of a loss completion unit. FIG. 1 is a diagram for explaining parameter estimation using a Gaussian process. A diagram showing updates using a Gaussian process, evaluation functions, and tables of estimated parameter values. FIG. 13 is a diagram showing an example of the configuration of a computer to which the present technique is applied.

Below, an embodiment of the present invention will be described. In the drawings used in the following description, components having the same functions and steps performing the same processing will be given the same reference numerals, and duplicate explanations will be omitted. In the following description, processing performed on each element of a vector or matrix will be applied to all elements of that vector or matrix, unless otherwise specified.

<Key Points of the First Embodiment>
(1) The reflectance of the wall surface is estimated so that the difference between a value based on the reverberation of an acoustic signal emitted from a specified position in a simulation space that simulates a specified space and a value based on the reverberation of an acoustic signal emitted from a position in the specified space, which is real space, that corresponds to the specified position in the simulation space, is minimized.

(2) By using RT60 as a value based on the reverberation of the acoustic signal, the effects of the position of the sound source and the microphone are reduced compared to the impulse response, making it possible to reduce the amount of calculations.

<Estimation system according to the first embodiment>
FIG. 1 is a functional block diagram of an estimation system according to a first embodiment, and FIG. 2 shows a process flow thereof.

The estimation system includes an acquisition unit 110, an estimation device 100, and an impulse response output unit 190.

The estimation system acquires point cloud data of the space where sound is collected via the acquisition unit 110, estimates the reflectance of a specified space using the point cloud data, and estimates and outputs the impulse response between an arbitrary sound source position and an arbitrary microphone position in the specified space using the estimated reflectance value i.

The estimation device 100 is a special device configured by loading a special program into a publicly known or dedicated computer having, for example, a central processing unit (CPU), a main storage device (RAM), etc. The estimation device 100 executes each process under the control of the central processing unit, for example. Data input to the estimation device 100 and data obtained in each process are stored in, for example, the main storage device, and the data stored in the main storage device is read out to the central processing unit as necessary and used for other processes. At least a part of each processing unit of the estimation device 100 may be configured by hardware such as an integrated circuit. Each storage unit of the estimation device 100 may be configured by, for example, a main storage device such as a RAM (Random Access Memory), or middleware such as a relational database or a key-value store. However, each storage unit does not necessarily need to be provided inside the estimation device 100, and may be configured by an auxiliary storage device configured by a semiconductor memory element such as a hard disk, optical disk, or flash memory, and may be configured to be provided outside the estimation device 100.

Each part is explained below.

<Acquisition unit 110>
The acquisition unit 110 acquires point cloud data R of the space where sound is collected (S110) and outputs it. Fig. 3 is a functional block diagram of the acquisition unit 110, and Fig. 4 shows an example of the processing flow. For example, the acquisition unit 110 includes a space sensing unit 111, a noise removal unit 113, and a space model combination unit 115.

<Spatial sensing unit 111>
The spatial sensing unit 111 acquires point cloud data of the space where sound is collected (S111) and outputs it. For example, the spatial sensing unit 111 is composed of a LiDAR installed in the space where sound is collected, emits light toward an object, calculates the distance to the object from the time difference between the emission and reception of the reflected light, and calculates the direction of the object from the emission direction. This distance and direction are represented by point cloud data. Various conventional technologies can be used as the spatial sensing technology. For example, Reference 1 is known as an existing spatial sensing technology.

(Reference 1) Toshio Ito, "Principles and Utilization of LiDAR Technology for Autonomous Driving", Science Information Publishing Co., Ltd., 2020 <Noise Removal Section 113>
The noise removal unit 113 receives the point cloud data, removes noise contained in the point cloud data (S113), and outputs the point cloud data after the noise removal. As the noise removal technique, various conventional techniques can be used. For example, Reference 2 is known as an existing noise removal technique.

(Reference 2) Rusu, RB, ZC Marton, N. Blodow, M. Dolha, and M. Beetz, "Towards 3D Point Cloud Based Object Maps for Household Environments", Robotics and Autonomous Systems Journal, 2008.
5 and 6 show examples of point cloud data after noise removal. Fig. 5 shows point cloud data acquired with the spatial sensing unit 111 (e.g., LiDAR) at an elevation angle of 0 degrees, and Fig. 6 shows point cloud data acquired with an elevation angle of -90 degrees. It is.

<Spatial Model Combination Unit 115>
The spatial model combining unit 115 receives a plurality of point cloud data, combines the plurality of point cloud data to restore the shape of the space (S115), and outputs point cloud data R indicating the shape of the space. For example, while rotating the LiDAR, a plurality of point cloud data with different elevation angles is obtained, and a three-dimensional scene is reconstructed by combining the plurality of point cloud data by an iterative closest point (ICP) algorithm. Various conventional techniques can be used as the spatial model combining technique. For example, Reference 3 is known as an existing spatial model combining technique.

(Reference 3) Szymon.R and Marc.L, "Efficient Variants of the ICP Algorithm", Proceedings Third International Conference on 3-D Digital Imaging and Modeling, 2001, pp. 145-152.
<Estimation device 100>
The estimation device 100 receives point cloud data R indicating the shape of a space, estimates the reflectance of a given space using the point cloud data R, and outputs an estimated value i of the reflectance.

For example, the estimation device 100 includes a room shape creation unit 120, a parameter input unit 130, an impulse response estimation unit 140, a reverberation time calculation unit 160, a real environment impulse response acquisition unit 150, an evaluation function calculation unit 170 and a parameter estimation unit 180 (see Figure 1).

The estimation device 100 may or may not include the acquisition unit 110 in the computer constituting the estimation device 100. When the acquisition unit 110 is not included, the estimation device 100 includes an input interface for inputting point cloud data (not shown). When the acquisition unit 110 is included, processing S113 and processing S115, or processing S115, may be performed in the computer constituting the estimation device 100. In this case, an input unit (not shown) is provided in front of the part corresponding to the processing performed in the computer, and the input unit receives information based on the point cloud data of the space where sound is collected (the point cloud data itself output by the spatial sensing unit 111, or the point cloud data with noise removed output by the noise removal unit 113).

<Room shape creation unit 120>
The room shape creation unit 120 receives point cloud data R indicating the shape of the space, complements the shape of the space that does not appear in the point cloud data R (S120), and outputs the shape U of the complemented space.

Figure 7 shows a functional block diagram of the room shape creation unit 120, and Figure 8 shows an example of the processing flow.

For example, the room shape creation unit 120 includes a wall surface extraction unit 121 and a defect completion unit 123.

<Wall Surface Extraction Unit 121>
The wall surface extraction unit 121 receives point cloud data R indicating the shape of the space, and extracts and outputs the wall surfaces that form the space using the shape of the space (S121).

For example, the wall surface extraction unit 121 extracts the wall surfaces by acquiring the parameters of the walls that form the space using an algorithm that improves RANSAC (Random Sample Consensus) for wall surface extraction. An example is shown below.

(1) Divide the point cloud data R that shows the shape of the space based on the observation angle. For example, divide it into M directions on the horizontal plane.

(2) Specify the mth direction (ROI (region of interest)) from among the M divided directions, where m = 1, 2, ..., M.

(3) Randomly select three points a, b, and c from the point cloud belonging to the ROI.

(4) Calculate the coefficients of the equation ax+by+cz+d=0 of the plane containing the three points a,b,c from the three points a,b,c. For example, calculate the coefficients from the cross product of the vectors ab and ac.

(5) Calculate the distance between each point of the point group belonging to the ROI and the plane, and find the number of points _nq whose distance is equal to or less than a predetermined threshold, where q = 1, 2, ..., Q. For example, calculate the distance d by d = n(pa), where n is the normal vector of the plane that includes the three points a, b, and c, and p is each point of the point group belonging to the ROI.

(6) Repeat the above steps (3) to (5) Q times, and among the Q numbers n _q, the maximum number of points n _m and the coefficient corresponding to that number are set as the parameters of the direction m.

(7) Repeat the above steps (2) to (6) M times and select the direction with the maximum number of points from the M numbers n _m .

(8) Collect points whose distance from the plane indicated by the coefficient corresponding to the direction selected in (7) above is less than a threshold, and find the normal vector from the collected points. For example, the normal vector can be found using SVD (Singular Value Decomposition) or PCA (Principal Component Analysis) of the covariance matrix.

(9) Record the normal vector as the largest wall surface in the point cloud data.

(10) Remove the point cloud data belonging to the largest wall from the data and return to (1). Note that the point cloud data belonging to the largest wall represents the extracted wall.

In addition, when the ratio of the point cloud data belonging to the largest wall to the entire point cloud data R is equal to or less than a certain value, it is determined that there is no wall there. In other words, when the ratio of the number of points n _q that are equal to or less than the threshold value in (5) above in a certain direction is equal to or less than a certain value, it is determined that there is no wall there. Furthermore, since the point cloud data belonging to the largest wall is removed from the data in (10) above, the number of walls that can be extracted gradually decreases and automatically converges. In addition, with this configuration, not only walls and ceilings but also a certain degree of planar shape of the point cloud that is likely to have reflections is extracted, and as a result, a space that also contains furniture and equipment is simulated. For example, the process may be controlled to be repeated until the plane formed by the point cloud data belonging to the largest wall is equal to or less than a predetermined threshold value (for example, 20 cm square). The predetermined threshold value may be estimated according to the amount of processing.

<Damage Compensation Unit 123>
The defect completion unit 123 receives the wall surfaces that form a space, and if there are objects other than the walls in the space and the shape of part of the wall surfaces cannot be estimated, it estimates the shape of that part of the wall surfaces from the shapes of the other parts of the wall surfaces that have been estimated, completes the shape of the space using the estimated part of the wall surfaces (S123), and outputs the shape U of the space after completion.

As mentioned above, LiDAR uses the reflection of a laser to measure distance, so data is lost in areas where the laser cannot reach due to obstructions. Therefore, the missing data is supplemented using estimated wall information to find the vertices (room corners).

In the example of Figure 9, display D is an obstruction, and the wall surface in the area surrounded by the dashed line in the figure cannot be observed. Note that Figure 9 shows point cloud data on a horizontal plane with an elevation angle of 0 degrees, and point O is the installation position of the LiDAR.

For example, the parameters of the wall (a, b, c, d) are extracted by plane approximation, and the coordinates of the vertices of the wall are found from the straight lines. In other words, as shown in Figure 10, the intersection point P of the straight lines is found as the coordinates of the vertex of the wall, and the wall is treated as being located in the dashed line area in Figure 10.

<Parameter Input Unit 130>
The parameter input unit 130 receives the sound source position S and the microphone position M as input (S130), and outputs them to the impulse response estimation unit 140. A speaker and a microphone are placed at positions in the real space corresponding to the sound source position S and the microphone position M in a simulation space that simulates a predetermined space. Therefore, the sound source position S and the microphone position M are positions where the speaker and the microphone can be placed in the real space. One or more positions can be assumed as the sound source position S and the microphone position M, respectively. The parameter input unit 130 is, for example, an input device such as a keyboard or a mouse, and the sound source position is input by a user via the parameter input unit 130. For example, the parameter input unit 130 may receive the shape of the space after completion, which is the output of the loss completion unit 123 (shown by a dashed line in FIG. 1), and present the shape of the space after completion to the user via a display device such as a display, and the user may specify where to place the sound source in the space after completion using a mouse or the like. Alternatively, the point cloud data R acquired from the LiDAR may be automatically input using object recognition as described in Reference 4, or the recognized object may be manually selected from its position.

(Reference 4) D. Maturana and S. Scherer, "VoxNet: A 3D Convolutional Neural Network for real-time object recognition", 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Hamburg, 2015, pp. 922-928, doi: 10.1109/IROS.2015.7353481.
The parameter input unit 130 is a storage medium reading device, and the estimation device 100 receives the sound source position and the microphone position as input by reading the storage medium in which the sound source position and the microphone position are stored via the parameter input unit 130. Good too.

Alternatively, the sound source position and microphone position may be manually input into the estimation device 100 in advance.

<Impulse response estimation unit 140>
The impulse response estimation unit 140 receives as input the sound source position S, the shape of the space U, the listening position M, and the estimated reflectance value R or the initial value R _ini , and uses these values to estimate the impulse response of the sound source position S and the listening position M in the shape of the space U (S140), and outputs the estimated impulse response value i _e . The initial value R _ini is assumed to be input manually in advance. For example, R _ini =0.5. The estimated reflectance value R is the output of the parameter estimation unit 180, which will be described later.

For example, the impulse response estimation unit 140 calculates a spatial transfer function of the listening position M in the space at the sound source position S using the sound source position S, the shape U of the space, the estimated value R or the initial value R _ini of the reflectance, and the listening position M. Various conventional techniques can be used as the spatial transfer function calculation technique. For example, a simulation of sound wave propagation is performed from the shape U of the space and the estimated value R or the initial value R _ini of the reflectance, and a simulation is performed by the finite-difference time-domain method (FDTD) method to predict the arriving sound at the listening position M, and the spatial transfer function is calculated. Furthermore, the impulse response estimation unit 140 estimates the impulse response using the spatial transfer function.

<Real environment impulse response acquisition unit 150>
The real environment impulse response acquisition unit 150 collects and acquires an acoustic signal emitted from a position corresponding to a sound source position S at a position corresponding to a listening position M in a predetermined space, which is a real space, measures the impulse response at the position corresponding to the sound source position S and the position corresponding to the listening position M (S150), and outputs a measurement value i _m of the impulse response. For example, a TSP (Time Stretched Pulse) signal is reproduced from a speaker arranged at a position corresponding to the sound source position S in the predetermined space, which is a real space, and the sound is collected by a microphone arranged at a position corresponding to the listening position in the predetermined space, which is a real space, and the impulse response is measured. The microphone and the speaker may be separate devices, or may be an integrated device such as a smart speaker.

<Reverberation time calculation unit 160>
The reverberation time calculation unit 160 receives the impulse response estimate i _e and the impulse response measurement i _m as input, calculates the reverberation time _{r e of} the impulse response estimate i e and the reverberation time r _i of the impulse response measurement i _m (S160), and outputs them. For example, RT60 is used as the reverberation time. RT60 is the time it takes for the reverberation in a room to drop by 60 dB. RT60 is used because it does not change sensitively depending on the location or sound source position like the impulse response, and can be obtained robustly as a numerical value depending on the structure and materials of the room.

The combination of the impulse response estimation unit 140 and the reverberation time calculation unit 160 is also called the simulation acquisition unit, as it acquires a value (RT60) based on the reverberation of an acoustic signal emitted from a specified position (sound source position) in a simulation space that simulates a specified space.

In addition, the combination of the real environment impulse response acquisition unit 150 and the reverberation time calculation unit 160 is also referred to as a real space acquisition unit, because it acquires a value (RT60) based on the reverberation of an acoustic signal emitted from a position corresponding to a specified position in a specified space, which is real space.

<Evaluation function calculation unit 170>
The evaluation function calculation unit 170 receives the reverberation time r _e and the reverberation time r _m as input, calculates the evaluation value E using the difference between the reverberation time r _e and the reverberation time r _m (S170), and outputs it. The evaluation value can be, for example, the squared error between the RT60 of the estimated impulse response value i _e and the RT60 of the measured impulse response value i _m , and is expressed by the following equation.

E = (r _e -r _m ) ²
<Parameter Estimation Unit 180>
The parameter estimation unit 180 receives the evaluation value E as input, estimates the reflectance of the wall surface using the evaluation value (S180), and outputs an estimated value R.

In this embodiment, the parameter estimation unit 180 performs parameter estimation using a Gaussian process.

The items required for parameter estimation using the Gaussian process are the evaluation value E and the parameter to be manipulated. For example, the parameter to be manipulated is the reflectance of each wall surface.

Using observation points, the regression curve that follows a Gaussian distribution is narrowed down to the range of true values.

The Gaussian process is used to efficiently search for which points to obtain in order to find the maximum value.

Figure 11 is a diagram to explain parameter estimation using a Gaussian process. The vertical axis of the diagram represents the evaluation value, and the horizontal axis represents the reflectance index.

The parameter estimation unit 180 uniquely determines an evaluation function for each parameter, predicts a regression curve using a Gaussian distribution, and sequentially updates the predicted parameters (reflectance) that minimize the evaluation value E, making it possible to quickly search for the minimum value of the evaluation function.

If the number of iterations is less than or equal to a predetermined number (NO at S185), the parameter estimation unit 180 outputs the updated parameters to the impulse response estimation unit 140, performs processes S140 to S180, and repeats the parameter estimation.

If the number of iterations exceeds a predetermined number (YES in S185), the parameter estimation unit 180 outputs the parameters that result in the minimum evaluation function as the final estimated value _Rf of the reflectance.

<Impulse response output section 190>
The impulse response output unit 190 receives the sound source position S', the shape U of the space, the listening position M', and the final reflectance estimate _Rf , and uses these values to estimate the impulse response of the listening position M' in the space at the sound source position S' (S190), and outputs the impulse response estimate i. For example, the impulse response is estimated by a method similar to that of the impulse response estimation unit 140. Impulse responses at various positions can be estimated by changing the sound source position S' and the listening position M'.

The estimation device 100 may or may not include an impulse response output unit 190 within the computer constituting the estimation device 100.

＜Effects＞
The above configuration allows us to estimate the reflectance in a short time. In addition, by using RT60, the influence of the sound source position and microphone position is reduced compared to impulse response, making it possible to reduce the amount of calculation.

By applying this embodiment to a microphone and speaker included in a smart speaker, etc., and a device with a positioning function such as LiDAR, it is possible to estimate the exact structure of the surroundings and parameters such as the reflectance of the structure. Furthermore, it is possible to estimate the location of use and apply accurate simulation information to acoustic processing (noise suppression, reverberation suppression).

<Modification>
In this embodiment, the spatial shape U, which is the output value of the acquisition unit 110, is used as an input to the impulse response estimation unit 140, but BIM data (architectural CAD data) created when designing a building may be used as an input to the impulse response estimation unit 140. In this case, the acquisition unit 110 and the room shape creation unit 120 may not be included.

In this embodiment, RT60 is used as a value based on the reverberation of the acoustic signal, but an impulse response may also be used as a value based on the reverberation of the acoustic signal. In this case, the estimation device 100 does not need to include the reverberation time calculation unit 160. However, as described above, using RT60 reduces the influence of the position of the sound source and the position of the microphone compared to the impulse response, making it possible to reduce the amount of calculation.

In this embodiment, E = (r _e -r _m ) ² is used as the evaluation value, but other evaluation values may be used. For example, E = -(r _e -r _m ) ² may be used. In this case, the parameter estimation unit 180 outputs the parameters that result in the maximum evaluation function as the final reflectance estimate value _Rf .

In short, the difference between the value based on reverberation in the simulation space and the value based on reverberation in the real space is used, and (i) the larger the difference is, the larger the value that results in the smallest evaluation function is used as the final reflectance estimate _Rf , or (ii) the smaller the value that results in the largest evaluation function is used as the final reflectance estimate _Rf .

<Simulation results>
Since estimation using the FDTD method takes a very long time, this simulation uses the mirror image method. Pyroomacoustics is used for the mirror image method. The true value of the reflectance is set to 0.15, and the RT60 norm is used for the evaluation function instead of the L2 norm of the impulse response.

As a result of estimating one variable, it was confirmed that using the difference in RT60 is less accurate than directly calculating the difference in impulse response, but it does get closer to the true value. It was found that when filling in the gap with the actual measured value, RT60 has less error than impulse response.

Figure 12 is a table showing updates using the Gaussian process, the evaluation function, and the estimated parameter values. In Figure 12, iter represents the number of iterations, target represents the evaluation function, and var_abs represents the reflectance. The true value of 0.15 maximizes the evaluation function after 13 iterations. In this simulation, the evaluation function (squared error) is multiplied by a negative number to search for variables that maximize the evaluation function.

<Other Modifications>
The present invention is not limited to the above-mentioned embodiment and modified examples. For example, the above-mentioned various processes may be executed not only in chronological order as described, but also in parallel or individually depending on the processing capacity of the device executing the processes or as necessary. In addition, appropriate modifications are possible within the scope of the present invention.

<Program and recording medium>
The various processes described above can be implemented by loading a program that executes each step of the above method into the memory unit 2020 of the computer shown in FIG. 13 and operating the control unit 2010, input unit 2030, output unit 2040, etc.

The program describing the processing can be recorded on a computer-readable recording medium. Examples of computer-readable recording media include magnetic recording devices, optical disks, magneto-optical recording media, and semiconductor memories.

This program may be distributed, for example, by selling, transferring, lending, etc. portable recording media such as DVDs and CD-ROMs on which the program is recorded. Furthermore, this program may be distributed by storing it in a storage device of a server computer and transferring the program from the server computer to other computers via a network.

A computer that executes such a program, for example, first stores in its own storage device the program recorded on a portable recording medium or the program transferred from a server computer. Then, when executing a process, the computer reads the program stored in its own recording medium and executes the process according to the read program. As another execution form of this program, the computer may read the program directly from the portable recording medium and execute the process according to the program, or may execute the process according to the received program each time a program is transferred from the server computer to this computer. In addition, the server computer may not transfer the program to this computer, but may execute the above-mentioned process by a so-called ASP (Application Service Provider) type service that realizes the processing function only by issuing an execution instruction and obtaining the results. Note that the program in this embodiment includes information used for processing by an electronic computer that is equivalent to a program (such as data that is not a direct command to the computer but has a nature that specifies the processing of the computer).

In addition, in this embodiment, the device is configured by executing a specific program on a computer, but at least a portion of the processing content may be realized by hardware.

Claims (6)

An estimation device for estimating a reflectance of a wall surface in a predetermined space, comprising:
a simulation acquisition unit that acquires a value based on reverberation of an acoustic signal emitted from a predetermined position in a simulation space that simulates the predetermined space;
a real space acquisition unit that acquires a value based on reverberation of an acoustic signal emitted from a position corresponding to the predetermined position in the predetermined space, which is a real space;
a parameter estimation unit that estimates a reflectance of a wall surface by using an evaluation value that is a difference between a value based on reverberation in the simulation space and a value based on reverberation in the real space;
An estimation device comprising: The estimation device of claim 1,
an evaluation function calculation unit that calculates, as the evaluation value, a value based on a square error between a value based on reverberation in the simulation space and a value based on reverberation in the real space;
Estimation device. The estimation device according to claim 1 or 2,
The parameter estimation unit predicts a regression curve using a Gaussian distribution and estimates a reflectance at which the evaluation value is minimum or maximum.
Estimation device. The estimation device according to any one of claims 1 to 3,
The simulation acquisition unit includes:
an impulse response estimation unit that estimates impulse responses of a sound source position and a listening position in a simulation space that simulates the predetermined space;
a reverberation time calculation unit for calculating a reverberation time of an estimated impulse response;
The real space acquisition unit includes:
a real-environment impulse response acquisition unit that acquires an acoustic signal emitted from a position corresponding to the sound source position at a position corresponding to the listening position in the predetermined space, which is a real space, and measures impulse responses at the position corresponding to the sound source position and the position corresponding to the listening position;
a reverberation time calculation unit for calculating the reverberation time of the measured impulse response;
the value based on the reverberation in the simulation space is the reverberation time of an estimate of the impulse response;
the value based on reverberation in real space is the reverberation time of a measured impulse response;
Estimation device. An estimation method for estimating a reflectance of a wall surface in a predetermined space using an estimation device, comprising:
a simulation acquisition step in which the estimation device acquires a value based on reverberation of an acoustic signal emitted from a predetermined position in a simulation space that simulates the predetermined space;
a real space acquisition step in which the estimation device acquires, in the predetermined space, which is a real space, a value based on reverberation of an acoustic signal emitted from a position corresponding to the predetermined position;
a parameter estimation step in which the estimation device estimates a reflectance of a wall surface by using an evaluation value obtained by calculating a difference between a value based on reverberation in the simulation space and a value based on reverberation in the real space;
An estimation method including: A program for causing a computer to function as an estimation device according to any one of claims 1 to 4.

Images