Position estimation system and position estimation method

Abstract

To allow the accuracy of position estimation to be improved at low cost.SOLUTION: A radio wave transmitter transmits radio waves indicating a signal sequence, based on a predetermined communication standard. A receiver extracts frequency response data indicating a frequency response for each frequency band of the received radio waves, obtains a reception time of the radio waves, generates integrated data based on the frequency response data and auxiliary information related to the reception time, extracts from the integrated data an irregularity feature quantity that indicates irregularity over time in a waveform indicated by the frequency response of the received radio waves, and estimates a position of the receiver based on the irregularity feature quantity.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a location estimation system and a location estimation method. [Background technology]

[0002] In recent years, with the increasing number of systems that utilize multimodal data, user location information is an important source of information used in various systems. There are various positioning systems that acquire location information, and the Global Positioning System (GPS), which uses satellite radio waves, is a particularly popular method due to its ease of use. However, in indoor environments where satellite radio waves cannot reach, such as commercial facilities and factories, an alternative method to GPS requires the installation of dedicated devices for positioning and location estimation based on the radio waves they emit. Beacons, ultrasonic transmitters, and routers are all examples of such dedicated devices.

[0003] Generally, positioning methods using radio waves are vulnerable to noise such as radio wave interference from the environment. Due to the nature of radio wave paths, in addition to a straight path where there are no obstacles between the transmitter and the receiver, there are also reflective paths that pass through surrounding objects. As a result, when a radio wave is transmitted once from the transmitter, the receiver may receive multiple radio waves with different characteristics, and these multiple radio waves become noise that reduces positioning accuracy (a phenomenon called multipath fading).

[0004] Conventional methods require the dense installation of numerous transmitters within an environment to increase the number of LOS (Line of Sight) radio waves, which are free from obstacles between the transmitter and receiver. However, the greater the number of devices, the higher the overall maintenance costs become, and this high cost of continuous operation has been a barrier to the introduction of such systems. Furthermore, achieving LOS is difficult in environments such as construction sites, where it is difficult to secure space to install devices. Therefore, the industrial sector is seeking a method that can achieve indoor positioning using a small number of devices.

[0005] Fingerprinting, a positioning method that utilizes Wi-Fi (registered trademark) CSI (Channel State Information) data, is attracting attention as a method that can cover a wide area while limiting the number of devices installed in the environment. CSI is information that indicates the frequency response of received radio waves, and is generally expressed as waveform information that shows the amplitude and phase for each frequency band.

[0006] The Fingerprinting method defines a list of discrete locations in the environment, records fingerprint information (pairs of location information and observed CSI data) at each location, and creates a dataset. The Fingerprinting method compares the observed CSI data during location estimation with the dataset and predicts the location at which the CSI data is most likely to have been observed.

[0007] The Fingerprinting method can significantly reduce the number of devices required, as it can measure a wide area with a radius of approximately 100 m using a single commercially available router. Furthermore, the Fingerprinting method can be applied to environments where line of sight cannot be ensured, even in environments with many obstacles and a lot of interference, because it can build a data-driven predictor that includes interference data. This makes the Fingerprinting method a promising technology that can reduce maintenance costs and be applied to environments where it is difficult to install devices.

[0008] Non-Patent Document 1 and Non-Patent Document 2 are background technologies in this technical field. The technology described in Non-Patent Document 1 employs a prediction model using deep learning for position estimation using a fingerprinting method. When waveform data representing the amplitude and phase for each frequency band in CSI data for one time step is input to the prediction model, a classification probability for each measurement point is output. The technology described in Non-Patent Document 1 applies a one-dimensional convolutional neural network (CNN) to the waveform data in the prediction model, thereby extracting feature amounts in the frequency band direction and utilizing them for prediction.

[0009] The technology described in Non-Patent Document 2 inputs CSI data for 30 consecutive time steps, rather than one time step, into a prediction model using deep learning, with the aim of increasing the amount of information in the time direction. CSI data exhibits a waveform that is somewhat unique for each location, but noise that causes the waveform to fluctuate over time can appear. The technology described in Non-Patent Document 2 introduces a 2D CNN into the prediction model to quantify the CSI data, including its fluctuation characteristics. Specifically, the technology described in Non-Patent Document 2 collectively quantify the relationship between adjacent frequency bands within the 2D kernel used by the CNN and the time-evolving relationship between adjacent observations. [Prior art documents] [Non-patent literature]

[0010] [Non-Patent Document 1] Wang, Liping and Sudeep Pasricha. “A Framework for CSI-Based Indoor Localization with ID Convolutional Neural Networks.” 2022 IEEE 12th International Conference on Indoor Positioning and Indoor Navigation (IPIN), 2022, pp.1-8 [Non-patent document 2] Chen, Hao et al. “ConFi: Convolutional Neural Networks Based Indoor Wi-Fi Localization Using Channel State Information.” IEEE Access 5, 2017, pp.18066-18074 Summary of the Invention [Problem to be solved by the invention]

[0011] The waveforms indicated by CSI data change irregularly over time, i.e., multiple waveforms are observed irregularly at the same location. However, neither the technology described in Non-Patent Document 1 nor the technology described in Non-Patent Document 2 takes into account the irregularity of waveform changes, and this irregularity may cause a decrease in the accuracy of location estimation. Therefore, one aspect of the present invention improves the accuracy of location estimation at low cost. [Means for solving the problem]

[0012] In order to solve the above problem, one aspect of the present invention employs the following configuration: A position estimation system includes a radio wave transmitter and a receiver, wherein the radio wave transmitter generates a signal sequence and transmits radio waves indicating the signal sequence based on a predetermined communication standard, the receiver receives the radio waves, extracts from the radio waves frequency response data indicating a frequency response for each frequency band of the received radio waves, obtains a reception time of the radio waves, generates integrated data based on the frequency response data and auxiliary information related to the reception time, extracts from the integrated data an irregularity feature that indicates irregularity over time of a waveform indicated by the frequency response of the received radio waves, and estimates the position of the receiver based on the irregularity feature. [Effects of the Invention]

[0013] According to one aspect of the present invention, the accuracy of position estimation can be improved at low cost.

[0014] Problems, configurations, and effects other than those described above will become apparent from the following description of the embodiments. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a block diagram illustrating an example of a functional configuration of an indoor positioning system according to a first embodiment. [Figure 2] 2 is a block diagram showing an example of the hardware configuration of a radio wave transmitter and a receiver according to the first embodiment. FIG. [Figure 3] FIG. 3 is an explanatory diagram illustrating an example of a signal sequence measured by a receiver in the first embodiment. [Figure 4] FIG. 4 is an explanatory diagram illustrating an example of processing by a signal loss information adding unit in the first embodiment. [Figure 5] FIG. 10 is an explanatory diagram illustrating an example of processing by a time lag information adding unit in the first embodiment. [Figure 6] FIG. 4 is an explanatory diagram illustrating an example of processing by a position estimation unit in the first embodiment. [Figure 7] FIG. 4 is an explanatory diagram illustrating an example of processing by an irregularity feature quantity generating unit in the first embodiment. [Figure 8] FIG. 10 is a block diagram illustrating an example of a functional configuration of an indoor positioning system according to a second embodiment. [Figure 9] FIG. 10 is an explanatory diagram illustrating an example of processing by a sequence missing information adding unit in the second embodiment. [Figure 10] FIG. 10 is an explanatory diagram illustrating an example of processing by an order rearrangement information providing unit in the second embodiment. [Figure 11] 10 is an explanatory diagram illustrating an example of a determination result by a signal loss information assigning unit, a sequence loss information assigning unit, and a sequence change information assigning unit in the second embodiment. FIG. [Figure 12] FIG. 10 is an explanatory diagram illustrating an example of a process performed by a time delay information adding unit in the second embodiment. [Figure 13] FIG. 10 is an explanatory diagram illustrating an example of a process for calculating the difference between a sending time and a receiving time in the second embodiment. [Figure 14] FIG. 11 is a block diagram illustrating an example of a functional configuration of an indoor positioning system according to a third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0016] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. The following description and drawings are examples for explaining the present invention, and some omissions and simplifications have been made as appropriate for clarity of explanation. The present invention can be implemented in various other forms. Unless otherwise specified, each component may be singular or plural.

[0017] In the following explanation, various types of information may be described using expressions such as "table" and "list," but the various types of information may also be expressed in other data structures. To indicate that the information is not dependent on the data structure, "XX table," "XX list," etc. may be referred to as "XX information." When describing identification information, expressions such as "identification information," "identifier," "name," "ID," and "number" are used, but these are interchangeable.

[0018] In the drawings for explaining the embodiments, the same components are given the same names and symbols, and repeated explanations thereof may be omitted.

[0019] When using the aforementioned fingerprinting method in the field, location estimation must be performed based on data measured at various locations. However, the training data for the prediction model contains only a limited number of training locations. Therefore, predictions for other test locations must be made by analogy with the experience learned at the training locations.

[0020] In the fingerprinting method using CSI (Channel State Information) data, due to the characteristics of CSI data, generalization to interpolation points within the distribution range of training points is possible. The characteristics of this CSI data are that waveform patterns are similar between nearby points and have spatial continuity. That is, for example, if a measurement is performed at point M, midway between points A and B, 2 m apart, a waveform pattern intermediate between points A and B is observed. Therefore, for example, the technology described in Non-Patent Document 1, which predicts the classification probability for each training point, outputs a classification probability of 50% for each of points A and B, and calculates a weighted average of the coordinates using the classification probability as a weight, resulting in a coordinate that is midway between points A and B. Therefore, the technology described in Non-Patent Document 1 can make predictions by analogy for interpolation test points adjacent to two or more RPs. Furthermore, the ability to accurately predict data from untrained points by analogy is also referred to as spatial generalization.

[0021] However, while the prediction model in the technology described in Non-Patent Document 1 can demonstrate spatial generalization for data under low-noise LOS conditions, it has difficulty demonstrating spatial generalization under noisy NLOS (Non Line of Sight) conditions. This is due to the following two characteristics of measurement data under NLOS conditions.

[0022] The first characteristic is that when a receiver receives both direct-path radio waves and interference signals at the same location, there will be multiple representative data points that can be measured at each location. When multiple representative data points exist, for example, in the two-location example mentioned above, multiple waveform patterns may be observed at each of the two locations, and the intermediate waveform pattern cannot be uniquely determined. This significantly increases the total number of intermediate representation patterns across the entire environment, complicating the feature space within the learning model and making it difficult to extract effective features. To achieve generalization for a dataset with these characteristics, it is necessary to create features that take into account all waveform patterns that can be observed at each location and obtain intermediate representations within that feature space.

[0023] The technology described in Non-Patent Document 1 uses CSI data for one time step to perform location prediction, and therefore implicitly assumes that instantaneous measurements can fully represent the features of a location. However, for the reasons mentioned above, this assumption does not hold in environments with a lot of interference, and spatial generalization cannot be expected.

[0024] The second characteristic is that continuously measured CSI data contains information related to the frequency of occurrence in time-series data. Time-series data has two types of properties: time-evolution properties and frequency of occurrence properties. Time-evolution properties are properties related to the order in which each piece of data appears, and frequency of occurrence properties are properties related to how frequently each piece of data is observed. In general, recurrent neural networks (RNNs) and convolutional neural networks (CNNs), which are used to train time-series data, are designed to prioritize the time-evolution properties as features.

[0025] CSI data has the property that the waveforms of the direct path and the interference wave are measured while switching irregularly over time. Therefore, even if CSI data is measured continuously, there is no time-evolving relationship between adjacent data. On the other hand, since CSI data is mainly composed of multiple representative waveform patterns, it has the property of the occurrence frequency of each of the multiple representative waveform patterns.

[0026] For these reasons, commonly used RNNs and CNNs are unable to extract effective features from CSI time-series data, and do not acquire appropriate intermediate representations, making it difficult to achieve spatial generalization.

[0027] The technology described in Non-Patent Document 2 extracts features from CSI data of multiple time steps using CNN. However, because CNN is a model that assumes time-evolving properties as described above, it is difficult to extract appropriate features for the reasons described above, and spatial generalization cannot be expected.

[0028] The indoor positioning system of this embodiment makes it possible to extract effective features that can be expected to be spatially generalized in order to improve the positioning performance at test points of the Fingerprinting method.

[0029] In order to extract appropriate features that achieve spatial generalization, a prediction model is required that collectively converts the waveform pattern of CSI data, including interference waves, and the irregularity of the waveform pattern into features and utilizes them for position prediction. As mentioned above, CSI data, including interference waves, is observed with multiple waveforms that switch irregularly over time. For this reason, the indoor positioning system used in this experiment uses a model that converts occurrence frequency characteristics into features and makes predictions when CSI data from multiple times is input.

[0030] In addition to irregularities caused by interference waves, CSI data can also have irregularities in radio wave reception due to packet loss. Packet loss occurs due to surrounding network congestion and signal strength attenuation, and therefore shows different trends depending on the location. Furthermore, the likelihood of packet loss often shows similar trends between adjacent locations, making it information with spatial continuity. Therefore, the indoor positioning system of this embodiment integrates information indicating the frequency of packet loss with CSI data and uses it for prediction. Furthermore, the indoor positioning system of this embodiment adds metadata that can distinguish consecutive signals at the time of transmission, thereby characterizing the trends for each cause of packet loss (e.g., delayed reception, not being treated as received due to insufficient signal strength).

[0031] The indoor positioning system of this embodiment assigns metadata that distinguishes signals and uses a prediction model that characterizes their potential irregularities, making it possible to achieve highly accurate predictions even for test locations that are not included in the training data, even in environments with severe radio interference. [Example]

[0032] 1 is a block diagram showing an example of the functional configuration of an indoor positioning system, which is an example of a position estimation system. The indoor positioning system includes a radio wave transmitter 1 that transmits radio waves based on a predetermined communication standard, and a receiver 2 that estimates the position of a user by receiving and analyzing the radio waves transmitted by the radio wave transmitter 1. In this embodiment, an example will be described in which the radio waves based on the predetermined communication standard are Wi-Fi (registered trademark) radio waves (hereinafter simply referred to as radio waves).

[0033] The radio wave transmitter 1 is, for example, a router, which generates a signal sequence and transmits the generated signal sequence via radio waves. The radio wave transmitter 1 includes, for example, a signal sequence generator 10 that generates the signal sequence, and a signal transmitter 11 that executes processing for transmitting the signal sequence. The signal sequence generator 10 generates signal sequences that represent the same data. The signal transmitter 11 transmits the signals generated by the signal sequence generator 10 at predetermined time intervals.

[0034] The receiver 2 is, for example, a smartphone or a tablet terminal, and is carried by a user (who may be a human or a mobile object such as a robot) whose position is to be measured. The receiver 2 includes, for example, a signal receiving unit 20, a data integrating unit 21, a reception history recording unit 22, and a position estimating unit 23.

[0035] The signal receiving unit 20 receives a signal corresponding to the radio wave transmitted by the radio wave transmitter 1. The signal receiving unit 20 includes, for example, a CSI data extracting unit 200 and a reception time acquiring unit 201. The CSI data extracting unit 200 extracts CSI data from the signal corresponding to the radio wave received from the radio wave transmitter 1. The CSI data is an example of data indicating the frequency response (waveform information indicating the amplitude and phase for each frequency band) of the received radio wave for each frequency band. The reception time acquiring unit 201 acquires the time at which the radio wave is received. Note that in this embodiment, the time at which the radio wave is received may also be referred to as the signal reception time, focusing on the signal corresponding to the radio wave.

[0036] The data integrator 21 integrates CSI data (waveform data), reception time data (an example of auxiliary information), and metadata (an example of auxiliary information). The data integrator 21 includes, for example, a signal loss information assigner 210 and a time offset information assigner 211. The signal loss information assigner 210 estimates information related to signal loss due to packet loss and generates metadata indicating the estimated information. The time offset information assigner 211 estimates information related to a shift in the packet reception time outside the specified time and generates metadata indicating the estimated information. Details of the processing by the signal loss information assigner 210 will be described later using FIG. 4, and details of the processing by the time offset information assigner 211 will be described later using FIG. 5.

[0037] The reception history recording unit 22 stores data of previously received radio waves (a combination of CSI data, reception time data, and metadata) for a predetermined number of time steps as a history. The data integration unit 21, for example, integrates the history with radio wave data at the current time and uses it.

[0038] The location estimation unit 23 receives as input data in which the CSI data and metadata are integrated, and uses a machine learning model to estimate the location of the user carrying the receiver 2 and output the estimated location (coordinates). The location estimation unit 23 includes, for example, a waveform feature quantity generating unit 230, an irregularity feature quantity generating unit 231, a classification prediction unit 232, and a coordinate conversion unit 233.

[0039] The waveform feature quantification unit 230 converts waveform information of the CSI data into features. The irregularity feature quantification unit 231 converts potential trends in irregularities in the time axis direction of the received radio waves and the signals corresponding to the received radio waves into features. The classification prediction unit 232 performs classification prediction using the Fingerprinting method based on the extracted features. The coordinate conversion unit 233 converts the predicted classification probability into coordinate information. Details of the processing by the position estimation unit 23 will be described later using Figure 6.

[0040] 2 is a block diagram showing an example of the hardware configuration of each of the radio wave transmitter 1 and the receiver 2. Each of the radio wave transmitter 1 and the receiver 2 is configured by, for example, a computer 1000. The computer 1000 includes, for example, a CPU (Central Processing Unit) 1010, a memory 1020, an auxiliary storage device 1030, an input device 1040, an output device 1050, a communication device 1060, and an antenna 1070.

[0041] The CPU 1010 is an example of a processor, and executes programs stored in the memory 1020. The memory 1020 includes a ROM (Read Only Memory), which is a nonvolatile storage element, and a RAM (Random Access Memory), which is a volatile storage element. The ROM stores unchanging programs (e.g., a Basic Input / Output System (BIOS)). The RAM is a high-speed, volatile storage element such as a DRAM (Dynamic Random Access Memory), and temporarily stores programs executed by the CPU 1010 and data used when the programs are executed.

[0042] The auxiliary storage device 1030 is a large-capacity, non-volatile storage device such as a magnetic storage device (HDD (Hard Disk Drive)) or a flash memory (SSD (Solid State Drive)), and stores programs executed by the CPU 1010 and data used when the programs are executed. In other words, the programs are read from the auxiliary storage device 1030, loaded into the memory 1020, and executed by the CPU 1010.

[0043] The input device 1040 is a device such as a keyboard or a mouse that receives input from an operator. The output device 1050 is a device such as a display device or a printer that outputs the results of program execution in a format that can be viewed by the operator. Note that the computer 1000 does not necessarily have to include the input device 1040 or the output device 1050.

[0044] The communication device 1060 is a network interface device that controls communication (wireless communication and wired communication) with other devices in accordance with a predetermined protocol. The communication device 1060 may also include a serial interface such as a USB (Universal Serial Bus).

[0045] The communication device 1060 outputs a signal to the antenna 1070, and receives the signal from the antenna 1070. The antenna 1070 converts the signal output from the communication device 1060 into radio waves and radiates them into space, and converts the radio waves radiated into space into a signal and inputs it to the communication device 1060. For example, the signal transmitting unit 11 of the radio wave transmitter 1 outputs a generated signal to the communication device 1060 of the computer 1000 constituting the radio wave transmitter 1, the communication device 1060 outputs the signal to the antenna 1070, the antenna 1070 converts the signal into radio waves and radiates them into space. Also, for example, the antenna 1070 of the computer 1000 constituting the receiver 2 receives the radio waves radiated into space by the radio wave transmitter 1, converts the received radio waves into a signal and inputs it to the communication device 1060 of the computer 1000, and the communication device 1060 inputs the signal to the signal receiving unit 20 of the receiver 2.

[0046] A part or all of the programs executed by the CPU 1010 may be provided to the computer 1000 from a removable medium (such as a CD-ROM or flash memory) that is a non-transitory storage medium, or from an external computer equipped with a non-transitory storage device via a network, and may be stored in the non-volatile auxiliary storage device 1030 that is a non-transitory storage medium. For this reason, the computer 1000 may have an interface for reading data from removable media.

[0047] Each of the radio wave transmitter 1 and receiver 2 is a computer system configured on a single physical computer 1000 or on multiple logically or physically configured computers 1000, and may operate in separate threads on the same computer 1000, or may operate on a virtual computer constructed on multiple physical computer resources.

[0048] The CPU 1010 of the computer 1000 constituting the radio wave transmitter 1 includes, for example, a signal sequence generation unit 10 and a signal transmission unit 11, which are all functional units. The CPU 1010 of the computer 1000 constituting the receiver 2 includes, for example, a signal reception unit 20, a data integration unit 21, and a position estimation unit 23, which are all functional units.

[0049] The CPU 1010 of the computer 1000 constituting the radio wave transmitter 1 functions as a signal sequence generator 10 by operating in accordance with a signal sequence generation program loaded into a memory 1020 of the computer 1000, and functions as a signal transmitter 11 by operating in accordance with a signal transmission program loaded into the memory 1020. The relationships between programs and functional units are similar for each functional unit included in the CPU 1010 of the computer 1000 constituting the receiver 2.

[0050] In addition, some or all of the functions of the functional units included in the radio wave transmitter 1 and the functional units included in the receiver 2 may be realized by hardware such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field-Programmable Gate Array).

[0051] The auxiliary storage device 1030 of the computer 1000 constituting the receiver 2 provides, for example, a storage area for realizing the reception history recording unit 22. In addition, the auxiliary storage device 1030 of the computer 1000 constituting the receiver 2 stores a machine learning model used by the position estimation unit 23.

[0052] In this embodiment, some or all of the information stored in the auxiliary storage device 1030 of each computer 1000 may be stored in the memory 1020 of that computer 1000, or may be stored in an external database connected to that computer 1000.

[0053] It should be noted that in this embodiment, the information used by the indoor positioning system is not dependent on the data structure and may be represented in any data structure, for example, a table, a list, a database, or a queue as appropriate.

[0054] Fig. 3 is an explanatory diagram showing an example of a signal sequence measured by the receiver 2. Fig. 3 shows examples of signal sequences received by the receiver 2 when the radio wave transmitter 1 transmits radio waves carrying the same signal (signal A) at regular intervals (every time "0:01" has elapsed) and when the receiver 2 is located at a location with little interference (an environment where radio wave interference is unlikely to occur) and when the receiver 2 is located at a location with much interference (an environment where radio wave interference is likely to occur).

[0055] At point (i), where there is little interference, packet loss is unlikely to occur, so receiver 2 receives the same signal at the same time intervals as when it was transmitted by radio wave transmitter 1. On the other hand, at point (ii), where there is a lot of interference, packet loss is likely to occur, so receiver 2 is unable to receive the signal at certain times (at "0:02" and "0:06" after receiving the first signal).

[0056] 4 and 5, an example will be described in which the data integration unit 21 extracts metadata indicating irregularities in signal reception trends and integrates the metadata with CSI data. The data integrated by the data integration unit 21 is input to a deep learning model in the location estimation unit 23, so that individual signals are integrated into vector representations, and signals from multiple times are integrated into matrices.

[0057] 4 is an explanatory diagram showing an example of processing by the signal loss information adding unit 210. For example, the signal receiving unit 20 may set the reception time of the first signal received from the radio wave transmitter 1 as τ0 and the transmission interval as τ d Let the allowable error be τ ε Let τ d and τ ε is a predetermined value. At this time, the signal loss information adding unit 210 refers to the reception time acquired by the reception time acquiring unit 201, and calculates τ0+(n-1)τ for each natural number n that is 2 or more and is equal to or less than the number of predetermined time steps. d -τ ε <τ n <τ0+(n-1)τ d +τ ε The time τ that satisfies n If no signal is received at step n, it is determined that the nth signal is missing (packet loss occurred when the nth signal was received), and the determination result is reflected in the signal sequence.

[0058] The data combining unit 21 combines the CSI data extracted from each received signal with the time of the received signal. However, since no CSI data is extracted for the reception time when a signal is lost, the signal loss information adding unit 210 generates null CSI data (for example, CSI data in which all elements in a vector are 0) for each lost signal and adds the null CSI data to the reception time when the lost signal should have been received (for example, τ0+(m-1)τ when the mth signal is lost). d) are concatenated with each other. This allows the signal loss information assigning unit 210 to obtain input data in which combinations of CSI data and reception times are arranged in order of reception time. Note that the null CSI data is metadata generated by the signal loss information assigning unit 210.

[0059] 5 is an explanatory diagram showing an example of processing by the time lag information assigning unit 211. For each received signal, the time lag information assigning unit 211 associates time information (here, the difference obtained by subtracting the previous reception time from the current reception time) since the previous signal was received with the CSI data. The time information may be a continuous scalar value or a discretized vector representation (for example, a three-dimensional vector representing any of "0.1 seconds or less," "0.2 seconds or less," or "0.3 seconds or less" in increments of 0.1 seconds).

[0060] Furthermore, the time offset information assigning unit 211 may include, as the time information, information on the difference between the current reception time and each of a plurality of times, such as not only the difference obtained by subtracting the previous reception time from the current reception time, but also the difference obtained by subtracting the reception time before last from the current reception time. Note that this time information is metadata generated by the time offset information assigning unit 211. Note that since there is no previous reception time for the first reception time, metadata indicating this difference is not created.

[0061] 6 is an explanatory diagram showing an example of processing by the location estimation unit 23. The data integrator 21 outputs CSI data and metadata for each signal to the location estimation unit 23, and the location estimation unit 23 receives as input a matrix representation 601 in which vector representations of the CSI data and metadata for each signal are concatenated for T time steps. The matrix representation 601 is two-dimensional data consisting of an axis in the time direction and an axis in the CSI frequency band direction. The matrix representation 601 is an example of integrated data generated by the data integrator 21.

[0062] The waveform feature generating unit 230 applies N one-dimensional convolution layers to the input matrix representation. At this time, the waveform feature generating unit 230 applies convolution in the frequency band direction of CSI to extract waveform features for each signal. As a result of the convolution by the waveform feature generating unit 230, a three-dimensional feature having a "channel axis" in addition to a "time axis" and a "frequency band axis" is output and input to the irregularity feature generating unit 231.

[0063] Fig. 7 is an explanatory diagram showing an example of processing by the irregularity feature quantity generating unit 231. The processing by the irregularity feature quantity generating unit 231 will be described with reference to Fig. 6 and Fig. 7. The irregularity feature quantity generating unit 231 extracts features in the time axis direction from three-dimensional feature quantities 701 having independent time information, which are output from the waveform feature quantity generating unit 230.

[0064] The irregularity feature quantity generating unit 231 extracts features from the three-dimensional feature quantity 701 using a one-dimensional time-invariant CNN having k filters, thereby obtaining feature quantities 702 corresponding to each of the k filters (k feature quantities 702 are obtained). A time-invariant CNN is a model that concatenates and outputs convolution results for each filter for input data of T time steps. The irregularity feature quantity generating unit 231 averages each of the k feature quantities 702 in the time direction, thereby obtaining k feature quantities 703 in the time direction that are specialized for the occurrence frequency of each signal within the T time steps. Here, the occurrence frequency of each signal includes, for example, the occurrence frequency of the waveform indicated by the CSI data and the occurrence frequency of the reception of a signal indicated by the metadata (for example, the occurrence frequency of signal loss and the occurrence frequency of signal reception intervals).

[0065] The frequency of waveform occurrence indicates the irregularity of the waveform over time, and the frequency of occurrence of signal reception indicated by metadata indicates the irregularity of signal reception trends over time. Therefore, by calculating feature 703 specialized for such occurrence frequency, irregularity feature generator 231 enables location estimation that takes these irregularities into account. Specifically, even if changes in the received waveform, losses in the received signal, delays in the received signal, etc. occur, it is possible to suppress a decrease in the accuracy of location estimation by receiver 2. Furthermore, suppression of a decrease in the accuracy of location estimation can be achieved without providing an additional device, i.e., at low cost.

[0066] The irregularity feature quantity generating unit 231 applies M one-dimensional convolutional layers to a vector formed by concatenating k features 703 specialized for occurrence frequency, thereby outputting a feature 704 containing information on occurrence frequency, which is input to the classification prediction unit 232.

[0067] Returning to the explanation of Fig. 6, the classification prediction unit 232 applies L fully connected layers to the features 704 input from the irregularity feature quantity generation unit 231, and then finally applies a Softmax layer to output a vector representation 602 of classification probabilities indicating the probability that the user is located at each predefined measurement point within the site, and inputs this to the coordinate transformation unit 233. The coordinate transformation unit 233 calculates a weighted average of the coordinates of the measurement points using the classification probabilities indicated by the vector representation 602 output by the classification prediction unit 232 as weights, and outputs the location estimation result as coordinates.

[0068] The machine learning model used by the position estimation unit 23 is a model that, when a matrix representation 601 in which vector representations of the CSI data and metadata of each signal are concatenated for T time steps is input, outputs a vector representation 602 of a classification probability. In other words, the machine learning model is a model that realizes processing by the waveform feature quantification unit 230 (N one-dimensional convolutional layers in the frequency band direction), the irregularity feature quantification unit 231 (one-dimensional time-invariant CNN, averaging in the time direction, and M one-dimensional CNN in the frequency band direction), and the classification prediction unit 232 (L fully connected layers and a softmax layer).

[0069] The machine learning model is created in advance by supervised learning using, as training data, a pair of a matrix representation 601 obtained in the same environment as actual operation and the position of the receiver 2 that received the signal from which the matrix representation 601 was calculated (the coordinates of the measurement point within the site used in actual operation).

[0070] The data integration unit 21 may not include at least one of the signal loss information assigning unit 210 and the time lag information assigning unit 211. That is, at least one of the metadata indicating a signal loss and the metadata indicating the signal reception interval may not be generated. If at least one of the metadata indicating a signal loss and the metadata indicating the signal reception interval is not generated, it is assumed that the machine learning model has been created using training data that does not include at least one of the metadata. [Example]

[0071] In the second embodiment, differences from the first embodiment will be mainly described, and explanations of similarities to the first embodiment will be omitted as appropriate. In this embodiment, the radio wave transmitter 1 assigns metadata to a signal sequence, and the receiver 2 uses the assigned metadata to generate metadata related to irregularities in signal reception, and performs location estimation taking the generated metadata into consideration.

[0072] 8 is a block diagram showing an example of the functional configuration of an indoor positioning system. The signal sequence generator 10 of the radio wave transmitter 1 of this embodiment includes a metadata assigner 100. The metadata assigner 100 assigns metadata to the signals to make it possible to distinguish between consecutive signals included in the signal sequence generated by the signal sequence generator 10. Since the metadata indicates the transmission order of the signals, assigning the metadata to the signals allows the receiver 2 to generate separate information such as information indicating whether a loss has occurred in the received signal and information indicating whether the signal has been received late.

[0073] The metadata assigning unit 100 includes, for example, an order information assigning unit 1001 and a transmission time assigning unit 1002. The order information assigning unit 1001 assigns metadata for distinguishing consecutive signals. Specifically, for example, the order information assigning unit 1001 can generate three different types of metadata A, B, and C, and repeatedly assigns A, B, or C in this order to the signals generated by the signal sequence generating unit 10, corresponding to the transmission order. Therefore, the metadata for distinguishing consecutive signals is also metadata indicating the transmission order.

[0074] For ease of explanation, hereinafter, a signal to which metadata A is assigned, a signal to which metadata B is assigned, and a signal to which metadata C is assigned will be referred to as signal A, signal B, and signal C, respectively. Through processing by the order information assigning unit 1001, signals repeated in the order of signal A, signal B, signal C, signal A, signal B, signal C, signal A, ... are transmitted from the radio wave transmitter 1. The transmission time assigning unit 1002 assigns information indicating the transmission time to the signals generated by the signal sequence generating unit 10 as metadata.

[0075] In the signal receiving unit 20 of the receiver 2 of this embodiment, the signal receiving unit 20 further includes a metadata extracting unit 202, and the data integrating unit 21 further includes a sequence missing information assigning unit 212, a sequence reordering information assigning unit 213, and a time delay information assigning unit 214. The metadata extracting unit 202 extracts the metadata assigned by the metadata assigning unit 100 from the received signal and inputs it to the data integrating unit 21.

[0076] The order loss information assigning unit 212 estimates information about signal loss and generates metadata indicating the estimated information based on information indicating the transmission order included in the metadata extracted by the metadata extracting unit 202. The order change information assigning unit 213 estimates information about change in the order in which signals are received based on information indicating the transmission order included in the metadata extracted by the metadata extracting unit 202 and generates metadata indicating the estimated information.

[0077] The time delay information assigning unit 214 estimates information relating to a delay in the time of signal reception based on information indicating the transmission time included in the metadata extracted from the received signal by the metadata extraction unit 202, and generates metadata indicating the estimated information. Details of the processing by the sequence loss information assigning unit 212 will be described later using Fig. 9, details of the processing by the sequence change information assigning unit 213 will be described later using Fig. 10, and details of the processing by the time delay information assigning unit 214 will be described later using Fig. 12.

[0078] In addition, the matrix representation 601 used for training the machine learning model in this embodiment and the matrix representation 601 input to the machine learning model by the position estimation unit 23 are assumed to be integrated with metadata generated by the order missing information assignment unit 212, the order reordering information assignment unit 213, and the time delay information assignment unit 214.

[0079] Fig. 9 is an explanatory diagram showing an example of processing by the sequence missing information assigning unit 212. In the example of Fig. 9, signals are transmitted from the radio wave transmitter 1 in the order of signal A, signal B, signal C, signal A, signal B, signal C, signal A, etc., but the receiver 2 does not receive the first signal C or the third signal A.

[0080] For example, when a signal is received, the sequence loss information assigning unit 212 determines whether there is a loss in the last three received signals including the received signal, based on metadata indicating the sequence information. Specifically, for example, if the last three received signals including the received signal include one each of signal A, signal B, and signal C, the sequence loss information assigning unit 212 determines that there is no sequence loss in the received signal, and if not, determines that there is a sequence loss in the received signal.

[0081] Also, for example, when the sequence loss information assigning unit 212 receives a signal and determines whether there is a loss in the last five received signals including the received signal, if the last five received signals including the received signal include the received signal twice, the signal that should have been received just before the received signal twice, and the signal that should have been received just before the received signal once (if the received signal is signal A, it includes signal A twice, signal C twice, and signal B once), it determines that there is no sequence loss in the received signal, and if not, it determines that there is a sequence loss in the received signal.

[0082] The sequence loss information assigning unit 212, for example, links a flag indicating the determination result of sequence loss (0 when it is determined that there is no sequence loss, and 1 when it is determined that there is a sequence loss) to the CSI corresponding to the received signal. The flag indicating the determination result of sequence loss is metadata generated by the sequence loss information assigning unit 212. Note that, for example, when determining loss using the most recent three received signals, the sequence loss information assigning unit 212 may not generate the flag for the first received signal and the second received signal because there are no received signals for the most recent three received signals, or may determine a predetermined value (for example, 0) as the value of the flag.

[0083] The sequence loss information assigning unit 212 may use a simple binary format or a one-hot vector format as the format of the data to be concatenated. The sequence loss information assigning unit 212 may also concatenate flags indicating the determination results of multiple time windows (e.g., a time window covering the past three packets and a time window covering the past five packets). The sequence loss information assigning unit 212 may also concatenate the metadata including the number of packets lost within the same time window.

[0084] Fig. 10 is an explanatory diagram showing an example of processing by the order reordering information assigning unit 213. In the example of Fig. 10, signals are transmitted from radio wave transmitter 1 in the order of signal A, signal B, signal C, signal A, signal B, signal C, signal A, etc., but the order in which signals B and C are received by receiver 2 the first time is reversed.

[0085] For example, when a signal is received, the order reordering information assigning unit 213 determines whether the order of the most recent three received signals, including the received signal, has been reordered based on metadata indicating the order information. Specifically, for example, if the correct order can be achieved by reordering the signals before the received signal (i.e., the signal one signal before the received signal and the signal two signals before the received signal) in the most recent three received signals, including the received signal, the order reordering information assigning unit 213 determines that the order has been reordered; otherwise, it determines that the order has not been reordered. For example, if the received signal is signal A, the correct order of the most recent three received signals is signal B, signal C, and signal A. However, if the most recent three received signals are signal C, signal B, and signal A, the correct order can be achieved by reordering signal C and signal B, and therefore the order is determined to have been reordered.

[0086] The order reordering information assigning unit 213, for example, links a flag indicating the determination result of order reordering (0 when it is determined that no order reordering has occurred, and 1 when it is determined that order reordering has occurred) to the CSI corresponding to the received signal. The flag indicating the determination result of order reordering is metadata generated by the order reordering information assigning unit 213. Note that when determining whether to reorder using the most recent three received signals, for example, the order reordering information assigning unit 213 may not generate a flag for the first received signal and the second received signal because there are no received signals for the most recent three received signals, or may determine a predetermined value (for example, 0) as the value of the flag.

[0087] The order reordering information assigning unit 213 may use a simple binary format or a one-hot vector format as the format of the data to be linked. Furthermore, the order reordering information assigning unit 213 may link flags indicating the respective determination results of multiple time windows (for example, a time window covering the past three steps and a time window covering the past five steps). Furthermore, the order reordering information assigning unit 213 may link the metadata by including the number of steps that have been reordered within the same time window.

[0088] 11 is an explanatory diagram showing an example of the determination results by the signal loss information assigning unit 210, the sequence loss information assigning unit 212, and the sequence reordering information assigning unit 213. In the example of FIG. 11, it is assumed that signals are transmitted from the radio wave transmitter 1 in the order of signal A, signal B, signal C, signal A, signal B, signal C, and signal A. In the "(1) normal pattern," the receiver 2 receives the signals without any loss or reordering occurring, and therefore the determination results by the signal loss information assigning unit 210, the sequence loss information assigning unit 212, and the sequence reordering information assigning unit 213 are "none" for all of the received signals.

[0089] In the "(2) slightly delayed pattern," receiver 2 received signal C, which should have been received at time "0:02," at a time that was outside the allowable error range for time "0:02," but which was earlier than the reception time of the next signal, signal A. Since there was no signal received at a time that was within the allowable error range for time "0:02," the determination result for time "0:02" by signal loss information assigner 210 is "yes." However, since no loss actually occurred and the reception order of the signals was not changed, the determination results by order loss information assigner 212 and order change information assigner 213 are "no" for all received signals.

[0090] In the "(3) non-arrival pattern," signal C, which should have been received by receiver 2 at time "0:02," is missing. Since there is no signal received at a time within the allowable error range for time "0:02," the result of the determination by signal loss information assigner 210 for time "0:02" is "yes." Furthermore, for example, signal C is not included in the last three received signals based on signal A received at time "0:03," and therefore the result of the determination by order loss information assigner 212 for at least this signal A is "yes." On the other hand, for any of the signals received by receiver 2, even if the last three received signals based on that signal are rearranged, the correct reception order will not be achieved, and therefore the result of the determination by order rearrangement information assigner 213 for all received signals is "no."

[0091] In the "(4) significantly delayed pattern," receiver 2 receives signal C, which should have been received at time "0:02," at a time that is not included in the allowable error range for time "0:02" and that is later than the reception time of signal A, which should have been received next after signal C. Since no signal was received at a time that is included in the allowable error range for time "0:02," the result of the determination by signal loss information assigner 210 for time "0:02" is "yes." Furthermore, for example, signal C is not included in the last three received signals based on signal A received at time "0:03," and therefore the result of the determination by sequence loss information assigner 212 for at least this signal A is "yes." Furthermore, for example, the order of the last three received signals based on signal B received at time "0:04" is signal A, signal C, signal B, and if signals A and C are swapped in this order, the correct order will be signal C, signal A, signal B, so the judgment result by the order swapping information assignment unit 213 for at least this signal B is "yes."

[0092] Fig. 12 is an explanatory diagram showing an example of processing by the time delay information assigning unit 214. In the example of Fig. 12, signal A is transmitted from radio wave transmitter 1 at 0.01 s intervals starting at time "0:00", and receiver 2 receives signal A without any loss at +0.01 s, +0.13 s, +0.21 s, +0.32 s, +0.45 s, +0.56 s, +0.68 s, etc., after the time "0:00". Note that the time delay information assigning unit 214 does not use information indicating the order in which signals are transmitted, and therefore, in Fig. 12, all received signals are described as signal A without distinction.

[0093] The time delay information assigner 214 calculates the difference between the transmission time and the reception time based on the reception time acquired by the reception time acquirer 201 and the transmission time included in the metadata extracted by the metadata extractor 202, and then connects the calculated delay time to the CSI data. A specific example of a method for calculating the difference will be described later with reference to FIG. 13. The difference may be expressed as a continuous scalar value or a discretized vector (for example, a three-dimensional vector representing any of "0.1 seconds or less," "0.2 seconds or less," or "0.3 seconds or less" in increments of 0.1 seconds).

[0094] 13 is an explanatory diagram showing an example of the process of calculating the difference between the transmission time and the reception time by the time delay information assigning unit 214. Here, it is assumed that the clock in the radio wave transmitter 1 and the clock in the receiver 2 are not synchronized, that is, the clock that is the information source of the transmission time and the clock that is the information source of the reception time are not synchronized.

[0095] The time delay information adding unit 214 calculates the transmission time t TX,0 (extracted by the metadata extraction unit 202 from the metadata included in the first signal), and the reception time t RX,0 (acquired by the reception time acquisition unit 201 when the first signal is received) is set as the reference time.

[0096] The time delay information adding unit 214 calculates the difference d between the n-th transmission time and the n-th reception time. n The nth received signal is transmitted at time t TX,n and the reception time t of the nth received signal. RX,n and a predetermined transmission interval τ of the signal. d Using and, d n =nτ d -((t RX,n -t RX,0 )-(t TX,n -t TX,0By setting the reference time in this way, the time delay information assigning unit 214 can reduce the influence of the difference between the clock of the radio wave transmitter 1 and the clock of the receiver 2, and can easily synchronize the times.

[0097] As described above, in the indoor positioning system of this embodiment, the occurrence frequencies of signal reception indicated by the metadata generated by the sequence missing information assigning unit 212, the sequence reordering information assigning unit 213, and the time delay information assigning unit 214 all indicate irregularities in the signal reception trends over time, so location information estimation can be performed taking this irregularity into account. Furthermore, location estimation is performed taking into account the metadata generated by the sequence missing information assigning unit 212, the sequence reordering information assigning unit 213, and the time delay information assigning unit 214. Therefore, even if a situation occurs in which a received signal is lost, the sequence of the received signal is reordered, or the difference between the transmission time and the reception time becomes large, a decrease in the accuracy of location estimation by the receiver 2 can be suppressed. Furthermore, suppression of a decrease in the accuracy of location estimation can be achieved without providing an additional device, i.e., at low cost.

[0098] The data integration unit 21 may not include at least one of the signal loss information assigning unit 210, the time lag information assigning unit 211, the sequence loss information assigning unit 212, the sequence reordering information assigning unit 213, and the time delay information assigning unit 214. That is, at least one of the metadata indicating a signal loss, the metadata indicating the signal reception interval, the metadata indicating the sequence loss determination result, the metadata indicating the sequence reordering determination result, and the metadata indicating the difference between the transmission time and the reception time may not be generated. In particular, since both the signal loss information assigning unit 210 and the sequence loss information assigning unit 212 generate metadata related to a signal loss, either one of them may not be included. Note that if at least one of the metadata indicating a signal loss, the metadata indicating the signal reception interval, the metadata indicating the sequence loss determination result, the metadata indicating the sequence reordering determination result, and the metadata indicating the difference between the transmission time and the reception time is not generated, it is assumed that the machine learning model has been created using training data that does not include the at least one piece of metadata. [Example]

[0099] In the third embodiment, differences from the second embodiment will be mainly described, and a description of the same points as in the second embodiment will be omitted as appropriate. In this embodiment, time synchronization is performed between the radio wave transmitter 1 and the receiver 2.

[0100] 14 is a block diagram showing an example of the functional configuration of an indoor positioning system. The signal transmitting unit 11 of the radio wave transmitter 1 of this embodiment includes a time synchronization unit 110. The signal receiving unit 20 of the receiver 2 of this embodiment further includes a time synchronization unit 203.

[0101] The time synchronization unit 110 and the time synchronization unit 203 perform processing to synchronize the clocks in the radio wave transmitter 1 and the receiver 2. By performing synchronization by the time synchronization unit 110 and the time synchronization unit 203, the time delay information assignment unit 214 can acquire more accurate time difference information by simply calculating the difference by subtracting the transmission time from the reception time, without performing the simple synchronization processing shown in Fig. 13 .

[0102] The present invention is not limited to the above-described embodiments, but includes various modifications. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those including all of the described configurations. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment, or to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

[0103] Furthermore, the above-described configurations, functions, processing units, processing means, etc. may be partially or entirely implemented in hardware, for example, by designing them as integrated circuits. The above-described configurations, functions, etc. may also be implemented in software, with a processor interpreting and executing a program that implements each function. Information such as the programs, tables, and files that implement each function can be stored in a memory, a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card, SD card, or DVD.

[0104] In addition, the control lines and information lines shown are those that are considered necessary for the explanation, and do not necessarily show all the control lines and information lines in the product. In reality, it can be assumed that almost all components are interconnected. [Explanation of symbols]

[0105] 1 radio wave transmitter, 2 receiver, 10 signal sequence generation unit, 11 signal transmission unit, 20 signal reception unit, 21 data integration unit, 22 reception history recording unit, 23 position estimation unit, 1000 computer, 1010 CPU, 1020 memory, 1030 auxiliary storage device, 1060 communication device, 1070 antenna

Claims

1. 1. A position estimation system, comprising: A radio wave transmitter and a receiver are provided, The radio wave transmitter is generating a signal sequence; transmitting radio waves representing the signal sequence based on a predetermined communication standard; The receiver includes: receiving the radio wave, extracting frequency response data indicating a frequency response for each frequency band of the received radio wave from the radio wave; Acquire the time when the radio wave was received; generating integrated data based on the frequency response data and the auxiliary information relating to the reception time; extracting, from the integrated data, an irregularity feature that indicates irregularity over time in a waveform that is indicated by the frequency response of the received radio wave; A position estimation system that estimates a position of the receiver based on the irregularity feature.

2. 2. The location estimation system of claim 1, The receiver includes: generating information relating to at least one of a loss of a signal indicated by the radio wave and a deviation in the time of reception of the signal based on the time of reception; Including the generated information in the auxiliary information; A position estimation system that extracts from the integrated data the irregularity feature that indicates irregularity over time in the waveform indicated by the frequency response of the received radio waves and the reception tendency of the received radio waves.

3. 3. The position estimation system of claim 2, The receiver includes: generating signal loss information indicating whether a loss has occurred in the signal indicated by the radio wave based on the reception time, the time when the radio wave was first received, a predetermined transmission interval of the radio wave by the radio wave transmitter, and a predetermined allowable range error; The signal loss information is included in the auxiliary information; A position estimation system that extracts, from the integrated data, the irregularity feature that indicates irregularities over time in the waveform indicated by the frequency response of the received radio wave and in signal loss.

4. 3. The position estimation system of claim 2, The receiver includes: generating information indicating a difference from a previous reception time for each of the reception times; information indicating the difference is included in the auxiliary information; A position estimation system that extracts from the integrated data the irregularity feature that indicates irregularity over time in the waveform indicated by the frequency response of the received radio waves and the difference in the reception time of the radio waves.

5. 3. The position estimation system of claim 2, the radio wave transmitter transmits the radio waves to which a transmission time has been assigned to each signal included in the signal sequence; The receiver includes: acquiring the transmission time from the radio wave; generating information indicating a difference between the reception time and the transmission time; information indicating the difference is included in the auxiliary information; A position estimation system that extracts from the integrated data the irregularity feature that indicates irregularity over time in the waveform indicated by the frequency response of the received radio waves and the delay in the reception time of the radio waves.

6. 6. The position estimation system of claim 5, The receiver calculates the difference based on the transmission time, the reception time, and a predetermined transmission interval of the radio waves by the radio wave transmitter, and using the first transmission time and the first reception time as references.

7. 6. The position estimation system of claim 5, Synchronizing the time between the radio wave transmitter and the receiver; The location estimation system calculates the difference by subtracting the transmission time from the reception time.

8. 2. The location estimation system of claim 1, the radio wave transmitter transmits the radio wave to which distinction auxiliary information for distinguishing successive signals is added to each signal included in the signal sequence; The receiver includes: acquiring the discrimination auxiliary information from the radio wave; generating information relating to at least one of a loss of a signal indicated by the radio wave and a change in the order of reception of the signal based on the differentiation auxiliary information; Including the generated information in the auxiliary information; A position estimation system that extracts from the integrated data the irregularity feature that indicates irregularity over time in the waveform indicated by the frequency response of the received radio waves and the reception tendency of the received radio waves.

9. 9. The location estimation system of claim 8, The receiver includes: generating sequence loss information indicating whether a loss of a signal indicated by the radio wave has occurred based on the distinction auxiliary information corresponding to a predetermined number of signals most recently received; The sequence-missing information is included in the auxiliary information; A position estimation system that extracts from the integrated data a waveform indicated by the frequency response of the received radio wave, a loss in the signal, and an irregularity feature that indicates irregularities over time.

10. 9. The location estimation system of claim 8, The receiver includes: generating order change information indicating whether a change in the order of reception of the signals indicated by the radio waves has occurred based on the distinction auxiliary information corresponding to a predetermined number of signals most recently received; The order change information is included in the auxiliary information, A position estimation system that extracts, from the integrated data, the irregularity feature that indicates irregularity over time in the waveform indicated by the frequency response of the received radio waves and in the reordering of the reception order of the signals.

11. 2. The location estimation system of claim 1, The receiver includes: a machine learning model including a time-invariant convolutional neural network, which receives the integrated data and outputs information indicating the position of the receiver; inputting the integrated data into the machine learning model; a position estimation system that extracts the irregularity feature that indicates an occurrence frequency of a waveform indicated by the frequency response of the received radio wave based on the time-invariant convolutional neural network included in the machine learning model.

12. A location estimation method by a location estimation system, comprising: the location estimation system includes a radio wave transmitter and a receiver; The position estimation method includes: the radio wave transmitter generates a signal sequence; the radio wave transmitter transmits radio waves indicating the signal sequence based on a predetermined communication standard; The receiver receives the radio wave, the receiver extracts, from the received radio waves, frequency response data indicating a frequency response for each frequency band of the radio waves; The receiver acquires the reception time of the radio wave, the receiver generates integrated data based on the frequency response data and the aiding information regarding the time of reception; the receiver extracts, from the integrated data, an irregularity feature that indicates irregularity over time of a waveform that is indicated by the frequency response of the received radio wave; The receiver estimates a position of the receiver based on the irregularity feature.

Images