Location identification device and location identification method

Abstract

To improve the accuracy of identifying the location of the object being positioned. [Solution] The positioning device (10) includes a first similarity calculation unit (23) that calculates a first similarity between a plurality of first feature quantities generated based on ambient sounds acquired in advance at a plurality of locations in the positioning target area and a second feature quantity generated based on ambient sounds acquired by the sound receiving device (1) at a first time; a first identification unit (24) that identifies a candidate first location where the object to be positioned (P) is located at a first time based on the first similarity; a second similarity calculation unit (25) that calculates a second similarity between a first feature quantity generated based on ambient sounds acquired in advance at each of the second locations that the object to be positioned (P) can reach at a second time, and a third feature quantity generated based on ambient sounds acquired by the sound receiving device (1) at a second time; and a second identification unit (26) that identifies the position of the object to be positioned (P) at a second time based on the second similarity.

Description

Technical Field

[0001] The present invention relates to a position specifying device and a position specifying method for specifying the position of a positioning target moving within a positioning target area.

Background Art

[0002] When an operator performs equipment inspection etc. at a site, a system for specifying the position of the operator (positioning target) in real time is required to quickly detect an accidental accident.

[0003] As a technique for specifying the position of a positioning target such as a person, GNSS (Global Navigation Satellite System) such as GPS (Global Positioning System) is known. However, in these systems, radio waves hardly reach indoors and it is difficult to use them for indoor position specification. Also, in specifying the position of a positioning target indoors, methods using radio waves such as Wi-Fi (registered trademark) and Bluetooth (registered trademark) are known. However, these technologies require installation of Wi-Fi (registered trademark) base stations and Bluetooth (registered trademark) beacons, which are radio wave transmission facilities for position specification, inside or around a structure. Also, depending on the size and structure of the structure where position specification is performed, it may be difficult to receive radio waves and the position specification accuracy may decrease. Therefore, a position specifying method that can be used indoors and does not require installation of radio wave transmission facilities is desired.

[0004] As such a position estimation technique, Patent Document 1 discloses a technique in which a terminal device calculates the similarity between environmental sound signature data generated based on measurement sound data recorded in an arbitrary closed space and a plurality of environmental sound signature data in closed spaces stored in a signature list storage unit in advance, and outputs, as information representing the estimation result of the position of the terminal device, the identification information of the closed space associated with the environmental sound signature data for which the calculated similarity is the maximum.

Prior Art Documents

Patent Documents

[0005] [Patent Document 1] Japanese Patent Publication No. 2016-31520 [Overview of the Initiative] [Problems that the invention aims to solve]

[0006] In the technology described in Patent Document 1, if there are multiple ambient sound signature data with the highest similarity, it may not be possible to accurately output the estimated location of the terminal device.

[0007] One aspect of the present invention aims to realize a positioning device and a positioning method that can improve the accuracy of determining the position of an object to be positioned. [Means for solving the problem]

[0008] To solve the above problems, a positioning device according to one aspect of the present invention is a positioning device that identifies the position of an object to be positioned as it moves within a positioning target area, comprising: a first similarity calculation unit that calculates a first similarity for each of the plurality of first feature quantities, which are generated based on each of the ambient sounds previously acquired at a plurality of positions within the positioning target area, and a second feature quantity generated based on ambient sounds acquired at a first time by a sound receiving device of the object to be positioned; and a first similarity calculation unit that identifies a candidate first position where the object to be positioned is located at the first time based on the plurality of first similarities. The system includes: a identification unit; a second similarity calculation unit that calculates a second similarity between (1) a plurality of positions to which the object to be positioned can reach from the candidate first position at a second time a predetermined time after the first time, and a first feature quantity generated from among the plurality of first feature quantities based on ambient sound acquired in advance at each of the second positions, and (2) a third feature quantity generated at the second time based on ambient sound acquired by the sound receiving device; and a second identification unit that identifies the position from which the ambient sound that generated the first feature quantity for which the second similarity is equal to or greater than a predetermined value was acquired as the position of the object to be positioned at the second time.

[0009] To solve the above problems, a location determination method according to one aspect of the present invention includes: a first similarity calculation step of calculating a first similarity for each of the plurality of first feature quantities, which is generated based on each of the ambient sounds previously acquired at a plurality of locations in the positioning target area, and a second feature quantity generated based on ambient sounds acquired at a first time by a sound receiving device of the object to be positioned; a first determination step of identifying a candidate first location where the object to be positioned is located at the first time based on the plurality of first similarities; and (1) from the first time to a predetermined The method includes: a second similarity calculation step in which a plurality of positions that the object to be positioned can reach from the candidate first position at a second time after a certain time interval are defined as second positions, and a second similarity calculation step in which a second similarity is calculated between (1) a first feature quantity generated based on ambient sound acquired in advance at each of the second positions and (2) a third feature quantity generated based on ambient sound acquired by the sound receiving device at the second time interval, and a second identification step in which the position at the second time interval is identified as the position of the object to be positioned at the second time interval, where the ambient sound that generated the first feature quantity for which the second similarity is equal to or greater than a predetermined value was acquired.

[0010] Each aspect of the present invention may be implemented by a computer, in which case a control program for the location identification device that enables the computer to implement the location identification device by operating the computer as each part (software element) of the location identification device, and a computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention. [Effects of the Invention]

[0011] According to one aspect of the present invention, the accuracy of identifying the position of the object to be positioned can be improved. [Brief explanation of the drawing]

[0012] [Figure 1] This is a block diagram showing the main components of a location identification system according to one embodiment of the present invention. [Figure 2] This figure shows an example of a spectrogram generated based on ambient sound acquired at a specific location within a waste treatment plant. [Figure 3] This flowchart shows an example of a processing method for a location identification method using a location identification device according to one embodiment of the present invention. [Modes for carrying out the invention]

[0013] One embodiment of the present invention will be described in detail below. Figure 1 is a block diagram showing the main components of a positioning system 100 including a positioning device 10 in this embodiment. The positioning system 100 is a system that identifies the position of a target to be measured P as it moves within a measurement target area. The measurement target area is not particularly limited as long as it is inside a building, and may be, for example, the inside of a factory such as a waste treatment plant, or the inside of a commercial facility such as a shopping mall.

[0014] (Configuration of location tracking system 100) As shown in Figure 1, the location identification system 100 includes a sound receiving device 1 and a location identification device 10.

[0015] The sound receiving device 1 acquires ambient sound, which is the sound around the sound receiving device 1. The sound receiving device 1 is located on the target P of the positioning device 10, and acquires ambient sound around the target P when the target P moves within the measurement target area. The sound receiving device 1 comprises a sound receiving unit 1a and an output unit 1b. The sound receiving unit 1a receives ambient sound, which is the sound around the sound receiving device 1. The output unit 1b outputs the ambient sound received by the sound receiving unit 1a to the positioning device 10. The sound receiving device 1 may be, for example, a smartphone. In the positioning system 100 of this embodiment, the sound receiving device 1 is configured to include both a sound receiving unit 1a and an output unit 1b, but the present invention is not limited thereto. In one aspect of the positioning system of the present invention, the system may be configured to include a sound receiving device that receives the above-mentioned ambient sound, and an output device that outputs the ambient sound received by the sound receiving device to the positioning device 10.

[0016] The position specifying device 10 is a device that specifies the position of the positioning target P moving within the measurement target area. The position specifying device 10 includes a communication unit 11, a control unit 20, a storage unit 12, an input unit 13 that receives an input to the position specifying device 10, and a display unit 14 for displaying various information.

[0017] The communication unit 11 communicates with the sound receiving device 1. Through the communication unit 11, the ambient sound received by the sound receiving device 1 is input to the position specifying device 10. The communication between the communication unit 11 and the sound receiving device 1 may be performed by wireless communication or by wired communication.

[0018] The storage unit 12 stores various data used by the position specifying device 10. In the storage unit 12, a plurality of feature amounts generated based on each of the ambient sounds acquired in advance at a plurality of positions in the positioning target area are stored. Hereinafter, the feature amount will be described as the first feature amount. The first feature amount is an index indicating the features of the ambient sound at each of the plurality of positions in the positioning target area. The positions where the ambient sound is acquired in advance may be one point or a plurality of points for each room included in the positioning target area.

[0019] In the position specifying system 100 in the present embodiment, a spectrogram which is a feature amount generated by continuously performing frequency analysis on the ambient sound over time and which represents the intensity of the frequency component with respect to time is used. That is, in the storage unit 12, a plurality of spectrograms generated based on each of the ambient sounds acquired in advance at a plurality of positions in the positioning target area are stored as the first feature amount.

[0020] FIG. 2 is a diagram showing an example of a spectrogram generated based on the ambient sound acquired at a certain position in the cleaning factory. In the spectrogram shown in FIG. 2, features peculiar to the ambient sound emitted from the devices arranged around the position where the ambient sound was acquired are shown.

[0021] The control unit 20 comprehensively controls each part of the position identification device 10. The control unit 20 includes an acquisition unit 21, a feature quantity generation unit 22, a first similarity calculation unit 23, a first identification unit 24, a second similarity calculation unit 25, and a second identification unit 26.

[0022] The acquisition unit 21 acquires the environmental sound received by the sound receiving device 1 that is output from the sound receiving device 1 and input to the position identification device 10 via the communication unit 11.

[0023] The feature quantity generation unit 22 generates a feature quantity of the environmental sound acquired by the sound receiving device 1 acquired by the acquisition unit 21. The feature quantity generation unit 22 generates a feature quantity of the same type as the first feature quantity for the environmental sound acquired by the sound receiving device 1. That is, in the present embodiment, the feature quantity generation unit 22 generates a spectrogram as the feature quantity.

[0024] The first similarity calculation unit 23 calculates the similarity between a plurality of first feature quantities (in other words, all the first feature quantities stored in the storage unit 12) stored in the storage unit 12 and a feature quantity (hereinafter referred to as the second feature quantity) generated by the feature quantity generation unit 22 based on the environmental sound acquired by the sound receiving device 1 possessed by the positioning target P at a certain time (hereinafter referred to as the first time) for each of the plurality of first feature quantities. Hereinafter, the similarity calculated by the first similarity calculation unit 23 is referred to as the first similarity. The first similarity calculation unit 23 may calculate, for example, the average value of the differences between the pixel values in a plurality of spectrograms generated based on each of the environmental sounds acquired in advance at a plurality of positions in the positioning target area and the pixel values in the spectrogram generated by the feature quantity generation unit 22 based on the environmental sound acquired by the sound receiving device 1 at the first time as the first similarity.

[0025] The first identification unit 24 identifies candidate locations for the positioning target P at a first time (hereinafter referred to as the first location) based on a plurality of first similarity values ​​calculated by the first similarity calculation unit 23. For example, the first identification unit 24 may identify a location as a candidate for the first location where an ambient sound was acquired that generated a first feature whose first similarity is greater than or equal to a predetermined value (in other words, an ambient sound used to generate a first feature whose first similarity is greater than or equal to a predetermined value). Alternatively, the first identification unit 24 may identify a location as a candidate for the first location where an ambient sound was acquired that generated a first feature whose first similarity is included in a predetermined rank from largest to smallest.

[0026] The second similarity calculation unit 25 calculates the similarity (hereinafter referred to as the second similarity) between a feature quantity (hereinafter referred to as the third feature quantity) generated by the feature quantity generation unit 22 based on ambient sound acquired by the sound receiving device 1 at a second time, a predetermined time after the first time, and a first feature quantity among a plurality of first feature quantities stored in the storage unit 12 that satisfies the following conditions. The second time quantity may be, for example, a time later than the first time quantity, for example, a time 1 to 10 seconds after the first time quantity. Here, the second position is defined as the position that the object to be positioned P can reach at the second time quantity from each of the candidate first positions identified by the first identification unit 24. For example, if the time between the first time quantity and the second time quantity is 5 seconds, the second position is defined as the position that the object to be positioned P can reach at the second time quantity from each of the candidate first positions 5 seconds after the first time quantity. The second similarity calculation unit 25 calculates the second similarity by determining the similarity between the first feature, which is generated based on ambient sounds previously acquired at each of the second positions, and the third feature, from among the multiple first feature quantities stored in the memory unit 12. The second similarity calculation unit 25 may, for example, calculate the second similarity by determining the average difference between the pixel values ​​in the multiple spectrograms generated based on ambient sounds previously acquired at each of the second positions, from among the multiple spectrograms stored in the memory unit 12, and the pixel values ​​in the spectrogram generated by the feature quantity generation unit 22 based on ambient sounds acquired by the sound receiving device 1 at the second time.

[0027] The second similarity calculation unit 25 may calculate the second similarity using a model trained by machine learning. That is, the second similarity calculation unit 25 may calculate the second similarity by inputting a third feature into a model for calculating the second similarity that has been generated in advance using multiple spectrograms generated based on each of the ambient sounds acquired in advance at multiple locations in the positioning target area. An example of a method for calculating the second similarity using machine learning is described below. However, the method for calculating the second similarity using machine learning is not limited to the method described below. In the method for calculating the second similarity using machine learning, first, for each of the multiple spectrograms generated based on each of the ambient sounds acquired in advance at multiple locations in the positioning target area, the spectrogram is treated as a matrix of pixel values ​​and an eigenvalue vector is calculated. At this time, if necessary, the calculated eigenvalue vector may be converted into a dimensionality-reduced vector using a method such as UMAP (Uniform Manifold Approximation and Projection). Next, a table data is created that links the coordinates of each of the multiple locations in the positioning target area with the eigenvalue vector. Next, the created table data is trained using a decision tree model such as Random Forest to create a model. Then, the second similarity score may be calculated by inputting the spectrogram generated by the feature generation unit 22 based on the ambient sound acquired by the sound receiving device 1 at the second time point, as a third feature, into the created model. This reduces the time required to calculate the second similarity score compared to calculating it using conventional algorithms.

[0028] The second identification unit 26 identifies the position of the object to be positioned P at a second time step based on the second similarity calculated by the second similarity calculation unit 25. The second identification unit 26 identifies the position of the object to be positioned P at a second time step by identifying the position where the ambient sound that generated the first feature quantity with a second similarity of a predetermined value or greater was acquired. For example, the second identification unit 26 may identify the position of the object to be positioned P at a second time step by identifying the position where the ambient sound that generated the first feature quantity with the maximum second similarity was acquired. Alternatively, if there are multiple first feature quantities with a second similarity of a predetermined value or greater, the second identification unit 26 may identify each position where the ambient sound that generated the first feature quantity was acquired as a candidate for the position of the object to be positioned P at a second time step. In one embodiment of the position identification system 100 of this disclosure, the second identification unit 26 may estimate the floor on which the object to be positioned P is located at a second time step based on the position information of the object to be positioned P at a second time step identified by the second identification unit 26.

[0029] (An example of the processing method for determining location using the location determination device 10) Next, an example of the location determination method processing by the location determination device 10 will be explained with reference to Figure 3. Figure 3 is a flowchart showing an example of the location determination method processing by the location determination device 10.

[0030] As shown in Figure 3, in the location determination method using the location determination device 10, first, the acquisition unit 21 acquires the ambient sound acquired by the sound receiving device 1 at the first time point from the sound receiving device 1 via the communication unit 11 (step S1). Next, the feature generation unit 22 generates feature quantities (i.e., second feature quantities) of the ambient sound acquired by the acquisition unit 21 (step S2).

[0031] Next, the first similarity calculation unit 23 calculates the similarity (i.e., first similarity) between the second feature generated by the feature generation unit 22 and the multiple first feature stored in the memory unit 12 for each of the multiple first feature (step S3, first similarity calculation step).

[0032] Next, the first identification unit 24 identifies candidate locations for the positioning target P at the first time step (i.e., the first location) based on a plurality of first similarity values ​​calculated by the first similarity calculation unit 23 (step S4, first identification step). Specifically, the first identification unit 24 identifies a location where ambient sound was acquired that generated a first feature quantity whose first similarity value is greater than or equal to a predetermined value as a candidate for the first location.

[0033] Next, the acquisition unit 21 acquires the ambient sound acquired by the sound receiving device 1 at a second time, a predetermined time after the first time, via the communication unit 11 (step S5). Next, the feature generation unit 22 generates feature quantities of the ambient sound acquired by the acquisition unit 21 (i.e., third feature quantities) (step S6).

[0034] Next, the second similarity calculation unit 25 calculates a second similarity between the first feature, which is the similarity between the first feature generated based on ambient sounds previously acquired at each of the second positions that the object to be positioned P can reach at the second time, from among the multiple first feature quantities stored in the storage unit 12, and the third feature (step S7, second similarity calculation step).

[0035] Next, the second identification unit 26 identifies the position of the object to be positioned P at the second time step based on the second similarity calculated by the second similarity calculation unit 25. Specifically, the position of the object to be positioned P at the second time step is identified by the location where the ambient sound that generated the first feature quantity with the maximum second similarity was acquired (step S8, second identification step).

[0036] As described above, the position identification device 10 in this embodiment includes: (1) a first similarity calculation unit 23 that calculates a first similarity between a plurality of first feature quantities stored in the memory unit 12 and a second feature quantity generated based on ambient sound acquired by the sound receiving device 1 at a first time; (2) a first identification unit 24 that identifies a candidate first position where the object to be positioned P is located at a first time based on the first similarity; (3) a second similarity calculation unit 25 that calculates a second similarity between a first feature quantity generated based on ambient sound acquired in advance at each of a plurality of second positions that the object to be positioned P can reach from the candidate first position at a second time, and a third feature quantity generated based on ambient sound acquired by the sound receiving device 1 at a second time; and (4) a second identification unit 26 that identifies the position where ambient sound was acquired that generated a first feature quantity whose second similarity is equal to or greater than a predetermined value as the position of the object to be positioned P at a second time.

[0037] According to this configuration, the second identification unit 26 identifies the position of the object to be positioned P at the second time step by targeting only the first feature quantity of the second position that can be reached from the first position identified based on the first similarity. Therefore, it is easier to identify the position of the object to be positioned P at the second time step compared to when all first feature quantities are used to identify the position of the object to be positioned P at the second time step. As a result, the accuracy of identifying the position of the object to be positioned can be improved. In addition, it is easier to narrow down the position of the object to be positioned P at the second time step to a single location compared to when all first feature quantities are used to identify the position of the object to be positioned P at the second time step. Furthermore, the amount of computation required to identify the position of the object to be positioned P at the second time step can be reduced compared to when all first feature quantities are used to identify the position of the object to be positioned P at the second time step. As a result, the time required to identify the position of the object to be positioned P can be reduced.

[0038] In the location identification system 100 of this embodiment, a spectrogram was used as the first, second, and third feature quantities, but the location identification system of this disclosure is not limited to this. In one aspect of the location identification system of this disclosure, the first, second, and third feature quantities may be Mel spectrograms. A Mel spectrogram is a feature quantity obtained by passing a spectrogram through a filter (also called a Mel filter bank) that has high resolution for low-frequency components and decreases resolution as the frequency increases, in accordance with the characteristics of human hearing, thereby converting the spectrogram to the Mel scale. In this case, the feature quantity generation unit 22 first generates a spectrogram based on the ambient sound acquired by the sound receiving device 1, and then generates a Mel spectrogram as a feature quantity by passing the spectrogram through the Mel filter bank. Generally, sound feature quantities such as spectrograms have a large number of dimensions and a large amount of information. Therefore, when sound feature quantities are used to identify a location, the amount of computation increases, and location identification takes time. In contrast, Mel spectrograms contain less information than spectrograms because they are generated by reducing the dimensionality of the spectrogram when it is passed through a Mel filter bank. Therefore, by using Mel spectrograms as the first, second, and third features, the computational load required for the first similarity calculation unit 23 and the second similarity calculation unit 25 to calculate the first and second similarity can be reduced. As a result, the time required to determine the position of the object P can be reduced.

[0039] In one embodiment of the location determination system 100 of the present invention, the second similarity calculation unit 25 may calculate a second similarity for each of the multiple candidates for the first location when the first determination unit 24 has determined multiple candidates for the first location. The second determination unit 26 may then determine the location where the ambient sound that generated the first feature quantity with the largest second similarity calculated for each of the multiple candidates for the first location was acquired as the location of the object to be determined P at the second time step. This allows the location of the object to be determined P at the second time step to be determined with high accuracy when the first determination unit 24 has determined multiple candidates for the first location.

[0040] [Examples of implementation using software] The function of the location identification device 10 (hereinafter referred to as "the device") is a program that causes the device to function as a computer, and can be realized by a program that causes the computer to function as each control block of the device (particularly each part included in the control unit 20).

[0041] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., memory) as hardware for executing the program. By executing the program using this control device and storage device, the functions described in each of the embodiments are realized.

[0042] The above program may be recorded on one or more computer-readable recording media, not temporary ones. These recording media may or may not be provided by the above device. In the latter case, the program may be supplied to the above device via any wired or wireless transmission medium.

[0043] Furthermore, some or all of the functions of each of the above control blocks can also be realized by logic circuits. For example, an integrated circuit in which logic circuits functioning as each of the above control blocks are formed is also included in the scope of the present invention. In addition, it is also possible to realize the functions of each of the above control blocks by, for example, a quantum computer.

[0044] The present invention is not limited to the embodiments described above, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0045] (summary) A positioning device according to aspect 1 of the present disclosure is a positioning device that identifies the position of a target to be positioned as it moves within a positioning target area, comprising: a first similarity calculation unit that calculates a first similarity for each of the plurality of first feature quantities, which are generated based on each of the plurality of first feature quantities, which are generated based on the ambient sounds acquired in advance at each of the plurality of positions within the positioning target area, and a second feature quantity, which is generated based on the ambient sounds acquired at a first time by a sound receiving device of the target to be positioned; and a first identification unit that identifies a candidate first position where the target to be positioned is located at the first time based on the plurality of first similarities, and (1 (1) A second similarity calculation unit that calculates a second similarity between (2) a first feature quantity generated based on ambient sound acquired in advance at each of the second positions, and (3) a third feature quantity generated based on ambient sound acquired by the sound receiving device at the second time, and (4) a position at the second time, where the position of the object to be positioned is determined to be a plurality of positions that the object to be positioned can reach from the candidate first position at a second time, a predetermined time after the first time, and (5) a plurality of positions where the object to be positioned can reach from the candidate first position, and (6) a first feature quantity generated based on ambient sound acquired in advance at each of the second positions, and (7) a third feature quantity generated based on ambient sound acquired by the sound receiving device at the second time, and (8) a position at the second time, where the position of the object to be positioned is determined to be a plurality of positions that the object to be positioned can reach from the candidate first position at a second time, a predetermined time after the first time.

[0046] The location identification device according to Embodiment 2 of this disclosure may be configured such that the first feature, the second feature, and the third feature are Mel spectrograms generated from the ambient sound, as described in Embodiment 1 above.

[0047] In the location identification device according to aspect 3 of this disclosure, the second similarity calculation unit may be configured to calculate the second similarity using a model trained by machine learning, as described in aspect 1 or 2 above.

[0048] The location identification device according to Embodiment 4 of the present disclosure may be configured such that, in Embodiment 1 or 2, the second similarity calculation unit calculates the second similarity for each of the multiple candidates for the first location when the first identification unit identifies a plurality of candidates for the first location, and the second identification unit identifies the location where the ambient sound that generated the first feature quantity with the largest second similarity calculated for each of the plurality of candidates for the first location was acquired as the location of the object to be positioned at the second time.

[0049] A positioning method according to aspect 5 of the present disclosure is a positioning method for identifying the position of an object to be positioned that is moving within a positioning target area, comprising: a first similarity calculation step of calculating a first similarity for each of the plurality of first feature quantities, which are generated based on each of the plurality of first feature quantities based on the ambient sounds previously acquired at each of the plurality of positions in the positioning target area, and a second feature quantity generated based on the ambient sounds acquired by a sound receiving device of the object to be positioned at a first time; and a first identification step of identifying a candidate first position where the object to be positioned is located at the first time based on the plurality of first similarities, and (1 (1) A second similarity calculation step, in which a plurality of positions that the object to be positioned can reach from the candidate first position at a second time a predetermined time after the first time are defined as second positions, and a second similarity is calculated between (2) a first feature quantity generated based on ambient sound acquired in advance at each of the second positions and (3) a third feature quantity generated based on ambient sound acquired by the sound receiving device at the second time, and (4) a second similarity calculation step, in which the position at the second time is identified as the position of the object to be positioned at the second time, in which the position at the second time is identified as the position at the second time, in which the ambient sound that generated the first feature quantity for which the second similarity is equal to or greater than a predetermined value was acquired. [Explanation of Symbols]

[0050] 1. Receiving device 10 Locating device 23 First similarity calculation unit 24 1st Specific Part 25 Second similarity calculation unit 26 Second Specific Part

Claims

1. A positioning device that identifies the position of an object to be positioned as it moves within a positioning target area, A first similarity calculation unit calculates a first similarity for each of the multiple first feature quantities, between a plurality of first feature quantities generated based on ambient sounds previously acquired at each of a plurality of locations in the positioning target area, and a second feature quantity generated based on ambient sounds acquired at a first time by a sound receiving device of the object to be positioned. A first identification unit identifies a candidate first position where the object to be positioned is located at a first time based on a plurality of first similarity values, (1) A second similarity calculation unit that calculates a second similarity between a plurality of positions that the object to be positioned can reach from the candidate first position at a second time a predetermined time after the first time, and a first feature quantity generated based on ambient sound acquired in advance at each of the second positions from among the plurality of first feature quantities, and (2) a third feature quantity generated based on ambient sound acquired by the sound receiving device at the second time. A positioning device comprising: a second identification unit that identifies the location where the ambient sound that generated the first feature quantity having a second similarity of a predetermined value or more was acquired as the position of the object to be positioned at the second time.

2. The location identification device according to claim 1, wherein the first feature, the second feature, and the third feature are Mel spectrograms generated from the ambient sound.

3. The location identification device according to claim 1 or 2, wherein the second similarity calculation unit calculates the second similarity using a model trained by machine learning.

4. When the first identification unit identifies a plurality of candidates for the first position, the second similarity calculation unit calculates the second similarity for each of the plurality of candidates for the first position. The position identification device according to claim 1 or 2, wherein the second identification unit identifies the position at the second time of the object to be positioned as the position of the object to be positioned, which is the position at the second time of the acquisition of the ambient sound that generated the first feature quantity that has the largest second similarity calculated for each of the plurality of candidate first positions.

5. A method for determining the position of an object to be positioned as it moves within a positioning target area, A first similarity calculation step, which calculates a first similarity for each of the multiple first feature quantities, between a plurality of first feature quantities generated based on ambient sounds previously acquired at each of a plurality of locations in the positioning target area, and a second feature quantity generated based on ambient sounds acquired at a first time by a sound receiving device of the object to be positioned. A first identification step of identifying a candidate first position where the object to be positioned is located at a first time based on a plurality of first similarity values, (1) A second similarity calculation step in which a plurality of positions that the object to be positioned can reach from the candidate first position at a second time a predetermined time after the first time are defined as second positions, and a first feature quantity generated based on ambient sound acquired in advance at each of the second positions from among the plurality of first feature quantities, and (2) a third feature quantity generated based on ambient sound acquired by the sound receiving device at the second time, A method for determining a location, comprising: a second determination step of determining the location where the ambient sound that generated the first feature quantity having a second similarity of a predetermined value or more was acquired as the location of the object to be positioned at the second time.

Images