Location estimation model construction device, location estimation device, location estimation model construction method, and program

Abstract

To automatically construct a position estimation model by using not only wireless LAN information with a label but also wireless LAN information without a label as learning data.SOLUTION: A position estimation model construction device includes: an information acquisition unit for acquiring wireless LAN information made of a reception signal intensity of each access point; and a position estimation model learning unit for learning a position estimation model for estimating the positional information from the wireless LAN information by using, as learning data, an aggregate of the wireless LAN information related to the positional information and the positional information and the wireless LAN information acquired by the information acquisition unit.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a technology for estimating a position using wireless LAN radio wave information. [Background technology]

[0002] Indoor position estimation in indoor environments where GPS is difficult to use is expected to have many applications, such as indoor navigation and monitoring, and is currently being actively researched.

[0003] Non-Patent Document 1 discloses a method using Wi-Fi (registered trademark) fingerprints, which is the most feasible indoor position estimation method. Note that hereinafter, wireless LAN technology such as Wi-Fi (registered trademark) will be referred to as "wireless LAN."

[0004] The method disclosed in Non-Patent Document 1 requires a large amount of training data that associates wireless LAN radio wave information with the indoor position coordinates where that information is observed, and therefore has a high implementation cost.

[0005] Meanwhile, in recent years, social networking services (SNS) that allow users to upload images taken with smartphone cameras have become widespread, and taking photos of our daily activities and uploading them to SNS has become commonplace in our daily lives. Furthermore, it has become easy to add contextual information such as GPS latitude and longitude (geotags) to photos taken with smartphones using EXIF. [Prior art documents] [Non-patent literature]

[0006] [Non-Patent Document 1] P. Davidson and R. Piche: A survey of selected indoor positioning methods for smartphones, IEEE Commu-nications Surveys Tutorials, vol.19, no.2, p.1347-1370, 2017, https: / / ieeexplore.ieee.org / document / 7782316 / . [Non-patent document 2] Zhang et al.: Location Semantic Labeling of Wi-Fi Information Using Convolutional Neural Networks and Camera Images, 69th Information Systems Research Meeting of the Institute of Electrical Engineers of Japan, IS-17-014 (March 2017). [Non-patent document 3] Diederik P. Kingma, Max Welling: Auto-Encoding Variational Bayes, ICLR, 2014, https: / / arxiv.org / abs / 1312.6114. Summary of the Invention [Problem to be solved by the invention]

[0007] A conventional technique for constructing an indoor position estimation system using wireless LAN radio wave information attached to a photograph is disclosed in, for example, Non-Patent Document 2. However, the conventional technique has a problem in that data without labels (information on indoor positions) cannot be used for model learning used in the indoor position estimation system.

[0008] The present invention has been made in consideration of the above points, and aims to provide a technology that enables automatic construction of a location estimation model by using unlabeled wireless LAN information as well as labeled wireless LAN information as training data. [Means for solving the problem]

[0009] According to the disclosed technology, an information acquisition unit that acquires wireless LAN information consisting of received signal strength for each access point; In designated facilities Wireless LAN information associated with location information , as a label A set of pairs of the location information and the information acquired by the information acquisition unit , unlabeled Using wireless LAN information as learning data, Acquired at the specified facility Estimate location information from wireless LAN information , neural network a location estimation model learning unit that learns a location estimation model, The set of pairs is a set obtained from a plurality of images with wireless LAN information. A location estimation model building device is provided. [Effects of the Invention]

[0010] According to the disclosed technology, a technology is provided that enables automatic construction of a location estimation model by using unlabeled wireless LAN information as well as labeled wireless LAN information as training data. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 is a configuration diagram of a location estimation model construction device 100. [Figure 2] 10 is a flowchart showing a processing flow. [Figure 3] FIG. 2 is a diagram illustrating the configuration of a position estimation device 200. [Figure 4] FIG. 10 is a diagram for explaining an outline of a store estimation procedure. [Figure 5] FIG. 1 is a diagram for explaining an overview of a position estimation model. [Figure 6] FIG. 10 is a diagram illustrating a network configuration of a sub-model. [Figure 7] FIG. 2 illustrates an example of a hardware configuration of the apparatus. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, an embodiment of the present invention (the present embodiment) will be described with reference to the drawings. The embodiment described below is merely an example, and the embodiment to which the present invention is applied is not limited to the following embodiment.

[0013] (Outline of the embodiment) In this embodiment, an indoor commercial facility such as a shopping mall is assumed, and a wireless LAN indoor position estimation system (position estimation model) in such an environment is automatically constructed using photos taken by a terminal such as a smartphone. However, each photo is assumed to have wireless LAN radio wave information observed by the smartphone when the photo was taken attached as metadata.

[0014] Since anyone can easily take photos using a smartphone, the information necessary to build a wireless LAN indoor positioning system can be easily collected using methods such as crowdsourcing.

[0015] Non-Patent Document 2 discloses a technique for automatically constructing a wireless LAN indoor position estimation system using wireless LAN radio wave information attached to such photographs. The technique disclosed in Non-Patent Document 2 uses a wireless LAN indoor position estimation method using linear discriminant analysis. On the other hand, this embodiment proposes an indoor position estimation method based on deep learning, which makes it possible to train a deep learning model by additionally using wireless LAN radio wave information without labels (indoor position coordinates).

[0016] In this embodiment, indoor position estimation is described as an example assuming an indoor commercial facility, but the application of the technology according to the present invention is not limited to this and can be applied to any location or facility. The technology according to the present invention can also be applied to outdoor position estimation.

[0017] (Environment assumed in this embodiment) As described above, in this embodiment, an indoor environment consisting of multiple stores, such as a shopping mall, is assumed. The floor map of the environment is assumed to be known. The floor map includes location information of each store (information on the area occupied by the store) and the store name. Furthermore, it is assumed that multiple wireless LAN access points are installed in the environment.

[0018] Furthermore, we assume that multiple camera images taken at multiple points within the store environment are given, where the nth camera image i n Wireless LAN information n is given. n is N A is an N-dimensional vector A is the number of access points in the environment, i.e., w n is a vector whose elements are the received signal strength of each access point. In other words, the w n An element in is the reception strength of a signal transmitted from a certain access point at the point where the camera image was captured. Hereinafter, "wireless LAN information" is assumed to be a vector whose elements are the reception signal strengths for each access point.

[0019] In addition, in this embodiment, a sequence of unlabeled wireless LAN information can be used as additional training data. This is assumed to be obtained from a smartphone or the like of a user walking in the environment, and is expressed as follows:

[0020] W u =[w1 u ,w2 u ,...,w t u ,...] However, w t u is the wireless LAN information obtained at time t.

[0021] (Device configuration, operation overview) In this embodiment, a location estimation model construction device 100 constructs (learns) a location estimation model. Fig. 1 shows an example of the configuration of the location estimation model construction device 100 in this embodiment.

[0022] 1, the location estimation model construction device 100 includes a store estimation unit 110, a store information DB (database) 120, an information acquisition unit 130, a location estimation model learning unit 140, and a location estimation model DB 150. An overview of the operation of the location estimation model construction device 100 will be described with reference to FIG.

[0023] In S1 (step 1), the store estimation unit 110 performs store estimation. The result of the store estimation is stored as store information in the store information DB 120. It is assumed that the store information DB 120 stores information necessary for learning a location estimation model in addition to the store information.

[0024] In S2, the location estimation model learning unit 140 reads out the store information from the store information DB 120 and the wireless LAN information W u The location estimation model is trained using the above. The trained location estimation model is stored in the location information model DB 150.

[0025] It should be noted that the "model" in this embodiment is a neural network model, and when stored in a storage unit such as a DB, it is stored as data consisting of weighting parameters and the like.

[0026] Furthermore, the location estimation model construction device 100 may not include the store estimation unit 110. In this case, store information obtained in advance is stored in the store information DB 120.

[0027] The position estimation model constructed (learned) by the position estimation model construction device 100 is used for position estimation using wireless LAN information. The configuration of a position estimation device 200 that performs such position estimation is shown in FIG.

[0028] As shown in FIG. 3, the position estimation device 200 includes an information acquisition unit 210, a position estimation unit 220, an output unit 230, and a position estimation model DB 240.

[0029] The position estimation model DB 240 stores trained position estimation models. The information acquisition unit 210 acquires wireless LAN information and inputs it to the position estimation unit 220. The position estimation unit 220 inputs the wireless LAN information to the position estimation model read from the position estimation model DB 240, thereby obtaining and outputting position information (information on the position where the wireless LAN information is measured) as an output from the position estimation model. The output unit 230 outputs the position information.

[0030] The location information output by the output unit 230 may be information indicating which point in the store the point at which the wireless LAN information was measured is (i.e., the store name, etc.), or it may be two-dimensional coordinates, or other information.

[0031] Note that by adding an output unit 230 to the position estimation model construction device 100, the position estimation model construction device 100 may be used as the position estimation device 200. Furthermore, the position estimation model construction device 100 may be called a position estimation device, and the position estimation model learning unit 140 may be called a position estimation unit.

[0032] The position estimation device 200 may be a device (for example, a server on a network) separate from the terminal that measures the wireless LAN information, or may be a terminal that measures the wireless LAN information.

[0033] Furthermore, the location estimation model learning unit 140 of the location estimation model construction device 100 and the location estimation unit 220 of the location estimation device have the same function in that they perform AP embedding processing (described later) and input wireless LAN information into the location estimation model to perform location estimation. The location estimation model learning unit 140 differs from the location estimation unit 220 in that it has a loss calculation function and a parameter update function.

[0034] The processing operations in the location estimation model construction device 100 will now be described in more detail.

[0035] (S1: Store Estimation) First, the store estimation process executed by the store estimation unit 110 will be described. In the store estimation process, it is estimated which store an image (an image with wireless LAN information) captured by a camera was captured at. Since the store estimation process executed by the store estimation unit 110 is itself a conventional technique disclosed in Non-Patent Document 2, only an outline of the process will be described. An outline of the process is shown in FIG. 4.

[0036] The store estimation process executed by the store estimation unit 110 is based on the idea that if a store in the environment is a branch of a chain store, its interior will likely be similar to other branches. Therefore, first, based on the store name of the store included in the environment, an image search engine is used to collect web images related to the interior of the chain store. These images are then used as training data to train a convolutional neural network (image classifier) ​​that estimates the store name from the captured image.

[0037] However, non-chain stores and stores for which not many web images were found are classified as "others." The class of each image taken with a smartphone or other device is then estimated using an image classifier. Because the location information of each store is included in the floor map, the approximate location where each image was taken can be obtained (except for images classified as "others"). Finally, outliers are removed using the wireless LAN information assigned to images classified as the same store.

[0038] The store estimation process in S1 makes it possible to obtain information that associates images, wireless LAN information, and stores (i.e., location information). This information is called store information. The store information is stored in the store information DB 120.

[0039] (S2: Location estimation model learning) Next, the location estimation model learning process executed by the location estimation model learning unit 140 will be described.

[0040] The position estimation model is a model that takes wireless LAN information as input and estimates the store where the wireless LAN information was observed. The position estimation model learning unit 140 learns the position estimation model by using the image (with the wireless LAN information attached) obtained in S1 and the information of the estimated store as learning data. The position estimation model of the present embodiment is composed of a plurality of neural network models (sub-models) in order to cope with the non-linearity of the wireless LAN space.

[0041] Fig. 5 shows an overview of the position estimation model. As shown in Fig. 5, each sub-model constituting the position estimation model takes wireless LAN information as input and outputs the position information of the point where the wireless LAN information was measured.

[0042] As will be described later, in the position estimation model, in order to cope with unstable wireless LAN signal strength, the correlation between the signal strengths of access points existing in the vicinity is utilized. Therefore, a convolutional neural network is used for each sub-model constituting the position estimation model to perform convolution processing on the received signal strengths of a plurality of access points. Therefore, in the input wireless LAN information (received signal strength vector), access points in the vicinity need to be located in adjacent dimensions.

[0043] Therefore, in S2, after the position estimation model learning unit 140 performs a process of determining the order of access points in the vector (referred to as "AP embedding"), the position estimation model (sub-model group) is learned.

[0044] <AP embedding> First, AP embedding will be described. The position estimation model learning unit 140 performs a process of determining the order of access points in the wireless LAN information (wireless LAN signal vector) as the AP embedding process. Assuming there are a plurality of wireless LAN signal vectors, the nth wireless LAN signal vector is represented as follows.

[0045] w n =[s n 1 ,s n2 ,…,s n j ,…,s n N_A )] T , However, s n j is the signal strength from the jth access point. As mentioned above, in the sequence of elements of the WLAN signal vector, elements corresponding to nearby access points must be located close to each other (e.g., jth and j+1th).

[0046] This problem can be considered as determining the order of access points given their physical proximity to each other. In this embodiment, the physical proximity (distance) between two access points is determined as the similarity (inverse of the correlation between signal strengths or average error) of the access points in the wireless LAN information assigned to the image group.

[0047] The location estimation model learning unit 140 of this embodiment solves this problem by applying it to the traveling salesman problem. The traveling salesman problem is a combinatorial optimization problem in which, given a set of cities and the travel costs between each pair of cities, one route is found that visits all cities exactly once and returns to the starting point with the smallest total travel cost.

[0048] The cities are access points, the distance (cost) between cities is the estimated physical distance between the access points, and the order of cities on the obtained route is w n This is used as the order of access points.

[0049] where c ab Let x be the distance between cities (access points) a and b. ab is a binary variable, and when it is 1, the route includes a path between cities a and b. When it is 0, it does not. The location estimation model learning unit 140 finds a route that minimizes the travel distance by solving the following optimization problem.

[0050]

number

[0051] Lines 3 and 4 are constraints to visit a city (access point) only once. Q represents a subset of cities (sub-route), and the constraint in line 5 is that the final route must visit all cities.

[0052] <Configuration of location estimation model> As described above, the location estimation model in this embodiment is composed of multiple sub-models. Each sub-model has its own area and is responsible for that area. Specifically, a sub-model is prepared for each store (area), and the sub-model is trained using training data obtained from that store and surrounding stores. In other words, the sub-model is responsible for that store and the surrounding stores. A surrounding store is defined as a store whose center is located within the top k nearest stores (k nearest stores) from the center of the target store. Here, there is no training data obtained from stores that are not chain stores (because the store estimation in S1 only estimates them as the "others" class). Therefore, for stores that are not chain stores, training is performed based on training data (wireless LAN information) virtually generated by the sub-models of the surrounding stores.

[0053] An example of the configuration of a submodel is shown in Figure 6. The submodel is designed based on a convolutional variational autoencoder (CVAE) (Non-Patent Document 3) and includes an encoder 10 and a decoder 20. Note that using a CVAE as a submodel is just one example. Autoencoders other than the CVAE may also be used as the submodel.

[0054] 6, the encoder 10 includes a Conv1D11, a Conv1D12, a Flatter 13, and a Dense 14. The decoder 20 includes a Dense 21, a Reshape 22, a DeConv1D23, and a DeConv1D24.

[0055] The input of the encoder 10 is wireless LAN information, and the wireless LAN information obtained by AP embedding in the order of access points is used as input. The encoder 10 outputs two-dimensional coordinates corresponding to the measurement positions of the input wireless LAN information, as well as the mean and variance (standard deviation) when the latent expression of the training data (wireless LAN information) is considered to be a normal distribution.

[0056] These two-dimensional coordinates are estimated coordinates of the location where the input Wi-Fi information was measured. CVAE can learn its parameters based on the constraint that the latent representation follows a normal distribution. This is because the distribution of signal strength in Wi-Fi information is assumed to follow a locally simple distribution. This assumption also makes it easier to generate virtual Wi-Fi information for non-chain stores.

[0057] The decoder 20 of the submodel restores the input wireless LAN information. Using this decoder 20, it is possible to (i) generate wireless LAN information from location coordinates and (ii) train a model from unlabeled data (wireless LAN information without location information).

[0058] <Learning the location estimation model> The location estimation model learning unit 140 learns a sub-model of the location estimation model using learning data (pairs of wireless LAN information and store (location information)) obtained from the store that the sub-model is responsible for and its surrounding stores. In learning, the location estimation model learning unit 140 updates the parameters of the sub-model (CVAE) so as to minimize the following function:

[0059]

number

[0060] The position estimation error is the error between the two-dimensional coordinate estimated by the encoder 10 and the correct data (correct store). If the two-dimensional coordinate is within the area of ​​the correct store, the error is 0; if not, the error is the distance between the two-dimensional coordinate and the boundary of the area of ​​the correct store that is closest to the two-dimensional coordinate.

[0061] The unlabeled WLAN information sequence observed in the vicinity of the area covered by the submodel (the W u ) is given, the data is used to train the submodel additionally. The series of wireless LAN information is assumed to be obtained while walking. From that series, the series of wireless LAN information that is close in Euclidean distance to the wireless LAN information used to train the submodel is used as additional training data. A close Euclidean distance means, for example, that the Euclidean distance is below a threshold. During training, the CVAE parameters are updated to minimize the following function:

[0062]

number

[0063] (Position estimation) The position estimation device 200 uses a trained position estimation model (a collection of sub-models) to estimate the store from which wireless LAN information has been collected. First, a sub-model for estimating the store from the wireless LAN information is selected. Here, it is assumed that the position estimation model DB 240 of the position estimation device 200 stores the trained position estimation model as well as the training data used in the training. In the following processing, the training data is also read from the position estimation model DB 240 and used.

[0064] That is, the location estimation unit 220 calculates the Euclidean distance between the wireless LAN information acquired by the information acquisition unit 210 and the average of the wireless LAN information used to train each sub-model, and the sub-model corresponding to the smallest distance is set as the main sub-model.Then, the sub-models of stores near the store in charge of the main sub-model are set as the sub-sub-models.

[0065] The location estimation unit 220 performs AP embedding processing on the wireless LAN information acquired by the information acquisition unit 210, and inputs the wireless LAN information after the AP embedding processing to the main sub-model and the sub-sub-model group. The location estimation unit 220 outputs the store that is closest to the average of the two-dimensional coordinates estimated by the respective encoders of the main sub-model and the sub-sub-model group as the final estimation result. Note that the average of the two-dimensional coordinates itself may also be output as the estimation result.

[0066] (Example of hardware configuration) The location estimation model construction device 100 and the location estimation device 200 described above can both be realized by, for example, causing a computer to execute a program. This computer may be a physical computer or a virtual machine on the cloud. The location estimation model construction device 100 and the location estimation device 200 are collectively referred to as "devices."

[0067] That is, the device can be realized by executing a program corresponding to the processing performed by the device using hardware resources such as a CPU and memory built into a computer. The program can be recorded on a computer-readable recording medium (such as a portable memory) and stored or distributed. The program can also be provided via a network such as the Internet or email.

[0068] Fig. 7 is a diagram showing an example of the hardware configuration of the computer. The computer in Fig. 7 includes a drive device 1000, an auxiliary storage device 1002, a memory device 1003, a CPU 1004, an interface device 1005, a display device 1006, an input device 1007, an output device 1008, and the like, all of which are interconnected by a bus BS.

[0069] A program for realizing processing on the computer is provided by a recording medium 1001 such as a CD-ROM or a memory card. When the recording medium 1001 storing the program is set in the drive device 1000, the program is installed from the recording medium 1001 to the auxiliary storage device 1002 via the drive device 1000. However, the program does not necessarily have to be installed from the recording medium 1001, but may be downloaded from another computer via a network. The auxiliary storage device 1002 stores the installed program as well as necessary files, data, etc.

[0070] The memory device 1003 reads and stores a program from the auxiliary storage device 1002 when an instruction to start the program is received. The CPU 1004 realizes the functions related to the device in accordance with the program stored in the memory device 1003. The interface device 1005 is used as an interface for connecting to a network, etc. The display device 1006 displays a GUI (Graphical User Interface) or the like according to the program. The input device 1007 is composed of a keyboard, mouse, buttons, a touch panel, etc., and is used to input various operation instructions. The output device 1008 outputs the results of calculations.

[0071] (Addendum) The following additional clauses are disclosed in relation to the above-described embodiment. (Additional note 1) Memory and at least one processor coupled to said memory; Including, The processor: Obtain wireless LAN information consisting of received signal strength for each access point, A set of pairs of wireless LAN information associated with location information and the location information, and the acquired wireless LAN information are used as learning data to learn a location estimation model that estimates location information from wireless LAN information. Location estimation model construction device. (Additional note 2) The processor determines the order of received signal strengths in the wireless LAN information consisting of received signal strengths for each access point by solving a traveling salesman problem in which cities are regarded as access points so that nearby access points are located in a close order, and inputs the wireless LAN information in the determined order into the location estimation model. Item 1. A location estimation model construction device according to claim 1. (Additional note 3) The localization model has multiple sub-models, each sub-model assigned to a different region. 3. The position estimation model construction device according to claim 1 or 2. (Additional note 4) Each of the sub-models constituting the plurality of sub-models has an encoder that receives wireless LAN information as an input and outputs location information, and a decoder that reconstructs the wireless LAN information input to the encoder. 4. A position estimation model construction device according to claim 3. (Additional note 5) The processor: The sub-model is trained using the set of pairs obtained from the area covered by the sub-model and its surrounding areas; Additional learning of the sub-model is performed using wireless LAN information observed in the vicinity of the area covered by the sub-model and acquired by the information acquisition unit. 5. The position estimation model construction device according to claim 3 or 4. (Additional note 6) A position estimation device that performs position estimation using a position estimation model learned by the position estimation model construction device according to any one of appended claims 1 to 5, Memory and at least one processor coupled to said memory; Including, The processor: Obtain wireless LAN information consisting of received signal strength for each access point, The acquired wireless LAN information is used as an input of the location estimation model to estimate location information of the location where the wireless LAN information was observed. Location estimation device. (Additional note 7) 1. A computer-implemented method for constructing a location estimation model, comprising: Obtain wireless LAN information consisting of received signal strength for each access point, A set of pairs of wireless LAN information associated with location information and the location information, and the acquired wireless LAN information are used as learning data to learn a location estimation model that estimates location information from wireless LAN information. Location estimation model construction method. (Additional note 8) A non-transitory storage medium storing a program executable by a computer to execute a location estimation model construction process, The location estimation model construction process includes: Obtain wireless LAN information consisting of received signal strength for each access point, A set of pairs of wireless LAN information associated with location information and the location information, and the acquired wireless LAN information are used as learning data to learn a location estimation model that estimates location information from wireless LAN information. Non-transitory storage medium.

[0072] Although the present embodiment has been described above, the present invention is not limited to such a specific embodiment, and various modifications and changes are possible within the scope of the gist of the present invention described in the claims. [Explanation of symbols]

[0073] 10 Encoders 20 Decoder 100 Location estimation model construction device 110 Store Estimation Department 120 Store Information DB 130 Information Acquisition Department 140 Location estimation model learning unit 150 Location Estimation Model DB 200 Position estimation device 210 Information Acquisition Department 220 Position estimation section 230 Output section 240 Location Estimation Model DB 1000 Drive Device 1001 Recording media 1002 Auxiliary storage 1003 Memory device 1004 CPU 1005 Interface device 1006 Display device 1007 Input Device 1008 Output Device

Claims

1. an information acquisition unit that acquires wireless LAN information consisting of received signal strength for each access point; a location estimation model learning unit that uses a set of pairs of wireless LAN information associated with location information in a predetermined facility and the location information as a label, and unlabeled wireless LAN information acquired by the information acquisition unit as learning data, to learn a location estimation model of a neural network that estimates location information from the wireless LAN information acquired in the predetermined facility, The set of pairs is a set obtained from a plurality of images with wireless LAN information. Location estimation model construction device.

2. The location estimation model learning unit determines the order of received signal strengths in the wireless LAN information consisting of received signal strengths for each access point by solving a traveling salesman problem in which cities are regarded as access points so that nearby access points are located in a close order, and inputs the wireless LAN information in the determined order into the location estimation model. The location estimation model construction device according to claim 1 .

3. The location estimation model has a plurality of sub-models, each sub-model being assigned to a different region of the given facility.

3. The position estimation model construction device according to claim 1.

4. Each of the sub-models constituting the plurality of sub-models is an autoencoder, and includes an encoder that receives wireless LAN information as input and outputs location information, and a decoder that reconstructs the wireless LAN information input to the encoder. The location estimation model construction device according to claim 3 .

5. The location information is information about an area in the specified facility, The position estimation model learning unit training the sub-model using the set of pairs obtained from the area covered by the sub-model and its surrounding areas; Additional learning of the submodel is performed using unlabeled wireless LAN information observed in the vicinity of the area covered by the submodel and acquired by the information acquisition unit.

5. The position estimation model construction device according to claim 3 or 4.

6. an information acquisition unit that acquires wireless LAN information consisting of received signal strength for each access point; a location estimation unit that uses, as learning data, a set of pairs of wireless LAN information associated with location information in a predetermined facility and the location information as a label, and unlabeled wireless LAN information acquired by the information acquisition unit, to learn a location estimation model of a neural network that estimates location information from the wireless LAN information acquired in the predetermined facility, and that estimates location information of a location where certain wireless LAN information is observed, using the learned location estimation model. The set of pairs is a set obtained from a plurality of images with wireless LAN information. Location estimation device.

7. 1. A computer-implemented method for constructing a location estimation model, comprising: Obtain wireless LAN information consisting of received signal strength for each access point, A neural network location estimation model is trained using a set of pairs of wireless LAN information associated with location information in a predetermined facility and the location information as a label, and the acquired unlabeled wireless LAN information as training data, to estimate location information from the wireless LAN information acquired in the predetermined facility. Location estimation model construction method.

8. A program for causing a computer to function as each unit in the location estimation model construction device according to any one of claims 1 to 5.

Images