Indoor positioning method and device based on RSSI fingerprints

Abstract

The embodiment of the invention provides an indoor positioning method and device based on RSSI (Received Signal Strength Indicator) fingerprints, and relates to the technical field of data processing, the method comprises the following steps: acquiring RSS (Received Signal Strength) data of WIFI (Wireless Fidelity) at an indoor reference point with known coordinates and at to-be-positioned positions on different time domains, setting a coefficient KS to extend fingerprint dimensions, and constructing an isomorphic RSSI fingerprint feature space; rSSI fingerprint features are reconstructed based on an attention mechanism, and the weight of important features is improved; extracting inherent hierarchical features in the fingerprint data based on a double-branch auxiliary positioning network, and fusing the inherent hierarchical features with the reconstructed features to improve the position discrimination of fingerprints; and a network is trained by using reference point information of known coordinates, and position coordinates of to-be-positioned fingerprint features in different time domains are predicted. By applying the scheme provided by the embodiment of the invention, the precision and robustness of the indoor fingerprint positioning system can be improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing technology, and in particular to an indoor positioning method and device based on RSSI fingerprint. Technical Background

[0002] Indoor positioning is one of the key technologies for location-based services. Various indoor positioning systems have been proposed, such as Ultra Wide Band (UWB), Wireless fidelity (WIFI), 5G, and Inertial Measurement Unit (IMU). Among them, fingerprint positioning has a huge advantage because it does not require base station location and target initial location. Fingerprint positioning is usually divided into offline and online phases: the offline phase detects the signal characteristics of all access points (APs) at all reference points (RPs), such as received signal strength indicator (RSSI) and channel state information (CSI), to form an offline fingerprint database; the online phase matches the fingerprint of the unknown location with the offline fingerprint database to predict the user's location.

[0003] With the prevalence of smart devices, fingerprint positioning based on Wi-Fi RSSI has been widely used in large multi-story buildings such as shopping malls and libraries, as it requires no additional equipment and is easy to collect, demonstrating significant application demand and market potential. However, in these application scenarios, due to the complex and diverse building structures and the dynamic and changing indoor environment, such as the possibility of base station equipment being removed, added, and replaced over a period of time, Wi-Fi RSSI-based fingerprint positioning systems struggle to obtain positioning signals with high location discrimination and long-term stability. This results in significant distribution discrepancies between the RSSI fingerprint database constructed in the offline phase and the online samples, leading to a sharp decline in system positioning accuracy. Consequently, related technologies face the following challenges: the feature spaces of fingerprints collected from the offline and online phases are heterogeneous and lack high location discrimination, making it difficult to obtain accurate matching results. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide an indoor positioning method and apparatus based on RSSI fingerprint, so as to increase the accuracy of an indoor wireless fingerprint positioning system and its robustness to changes in base station equipment.

[0005] In a first aspect, an embodiment of the present invention provides an indoor positioning method based on RSSI fingerprint, the method comprising:

[0006] Obtain the RSS data of Wi-Fi at the reference point with known indoor coordinates and the location to be located in different time domains, set the coefficient KS to expand the fingerprint dimension, and construct a homogeneous Wi-Fi RSSI fingerprint feature space;

[0007] The attention network is trained using the RSSI fingerprint at a reference point with known coordinates to obtain the attention matrix of the RSSI fingerprint at the location to be located in different time domains. Each RSSI fingerprint is reconstructed based on the attention matrix coefficient to increase the weight of important features.

[0008] The hierarchical feature extraction network is trained using the reconstructed RSSI fingerprint at the reference point with known coordinates. The hierarchical features of the RSSI fingerprint of the location to be located in different time domains are extracted and fused with the reconstructed features to improve the position discrimination of the fingerprint.

[0009] The positioning network is trained using the fused fingerprint at the reference point with known coordinates to obtain the position estimation of the fused fingerprint at the location to be positioned in different time domains.

[0010] In one embodiment of the present invention, the RSS data of the WIFI at the reference point with known indoor coordinates and the location to be located in different time domains are obtained, and the coefficient K is set. S Expand the fingerprint dimension and construct a homogeneous WiFi RSSI fingerprint feature space, including:

[0011] The RSS data of the WIFI base station equipment at the reference point with known coordinates and the location to be located in different time domains are obtained by mobile phone, the RSS data is normalized to obtain RSSI data, and then the coefficient K is used to calculate the RSSI data. S Expand the fingerprint dimension and obtain a homogeneous WIFI RSSI fingerprint feature space.

[0012] In one embodiment of the present invention, the RSSI fingerprint at a reference point with known coordinates is used to train an attention network, obtain an attention matrix of the RSSI fingerprint at the location to be located in different time domains, reconstruct each RSSI fingerprint based on the attention matrix coefficient, and increase the weight of important features, including:

[0013] Build an attention sub-network, which first calculates the attention matrix parameters w,b. Then calculate the attention matrix A based on the RSSI fingerprint. S ,A T , based on the attention matrix A S ,A T Reconstruct each RSSI fingerprint. Finally, introduce the Dropout layer to increase the robustness of the reconstructed fingerprint to the disappearance of APs.

[0014] In one embodiment of the present invention, the hierarchical feature extraction network is trained using the reconstructed RSSI fingerprint at a reference point with known coordinates, hierarchical features of the RSSI fingerprint of the location to be located in different time domains are extracted, and the hierarchical features are integrated with the reconstructed features to improve the position discrimination of the fingerprint, including:

[0015] A hierarchical feature extraction subnetwork is built. The subnetwork first uses the attention subnetwork to reconstruct the RSSI fingerprint features, then uses the feedforward neural network (FNN) to extract low-dimensional hierarchical features, and finally fuses the low-dimensional hierarchical features with the high-dimensional reconstructed features to obtain a fused fingerprint.

[0016] In one embodiment of the present invention, the method of training a positioning network using a fused fingerprint at a reference point with known coordinates to obtain position estimates of the fused fingerprint at the location to be positioned in different time domains includes:

[0017] A positioning network is built. The fused fingerprint obtained by splicing the attention subnetwork with the hierarchical feature subnetwork is first passed through a dimensionality-increasing FNN and then a dimensionality-reducing FNN to obtain a position estimate. Joint training is then performed, and the fused fingerprint at a reference point with known coordinates is input to train the positioning network. The position estimate of the fused fingerprint at the target location is obtained in different time domains.

[0018] In a second aspect, an embodiment of the present invention provides an indoor positioning device based on RSSI fingerprint, the device comprising:

[0019] Acquisition module: used to obtain the RSS data of WIFI at the reference point with known indoor coordinates and the location to be located in different time domains, and set the coefficient K S Expand the fingerprint dimension and construct a homogeneous RSSI fingerprint feature space.

[0020] Feature reconstruction module based on attention mechanism: The attention network is trained using the RSSI fingerprint at the reference point with known coordinates, and the attention matrix of the RSSI fingerprint of the location to be located in different time domains is obtained. Each RSSI fingerprint is reconstructed based on the attention matrix coefficient to increase the weight of important features.

[0021] Hierarchical feature extraction module: The hierarchical feature extraction network is trained using the reconstructed RSSI fingerprint at the reference point with known coordinates. The hierarchical features of the RSSI fingerprint of the location to be located in different time domains are extracted and fused with the reconstructed features to improve the position discrimination of the fingerprint.

[0022] Position prediction module: Use the fused fingerprint at the reference point with known coordinates to train the positioning network to obtain the position estimate of the fused fingerprint at the location to be positioned in different time domains.

[0023] In one embodiment of the present invention, the acquisition module is specifically configured to:

[0024] The RSS data of the WIFI base station equipment at the reference point with known coordinates and the location to be located in different time domains are obtained by mobile phone, the RSS data is normalized to obtain RSSI data, and then the coefficient K is used to calculate the RSSI data. S Expand the fingerprint dimension and obtain a homogeneous WIFI RSSI fingerprint feature space.

[0025] In one embodiment of the present invention, the feature reconstruction module based on the attention mechanism is specifically used to:

[0026] Build an attention sub-network, which first calculates the attention matrix parameters w,b. Then calculate the attention matrix A based on the RSSI fingerprint. S ,A T , based on the attention matrix A S ,A T Reconstruct each RSSI fingerprint. Finally, introduce the Dropout layer to increase the robustness of the reconstructed fingerprint to the disappearance of APs.

[0027] In one embodiment of the present invention, the hierarchical feature extraction module is specifically used to:

[0028] A hierarchical feature extraction subnetwork is built. The subnetwork first uses the attention subnetwork to reconstruct the RSSI fingerprint features, then uses the feedforward neural network (FNN) to extract low-dimensional hierarchical features, and finally fuses the low-dimensional hierarchical features with the high-dimensional reconstructed features to obtain a fused fingerprint.

[0029] In one embodiment of the present invention, the location prediction module is specifically configured to:

[0030] Build a positioning network, input the fused fingerprint of a reference point with known coordinates, first pass it through a dimensionality-increasing FNN and then through a dimensionality-reducing FNN to obtain a position estimate, and perform joint training. After obtaining this training model, input the fused fingerprint of the location to be located in different time domains to obtain the position estimate of the fingerprint to be located.

[0031] Beneficial effects of the embodiments of the present invention:

[0032] This invention provides an indoor positioning method and device based on RSSI fingerprints. While reconstructing the fingerprint feature space based on an attention mechanism, it also extracts and linearly fuses the hierarchical features of the fingerprint. These fused features are then used as fingerprints for indoor position estimation, contributing an important technical tool to the development of indoor location-based services. Compared to related technologies, this indoor fingerprint positioning method and device significantly improves the accuracy of fingerprint positioning in complex and dynamic indoor scenarios and its robustness to AP variations, increasing the feasibility of deploying fingerprint positioning systems in real-world indoor scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other embodiments can also be obtained based on these drawings.

[0034] Figure 1 A schematic diagram of a flow chart of an indoor positioning method based on RSSI fingerprint provided by an embodiment of the present invention;

[0035] Figure 2 A schematic diagram of the transformation from heterogeneous features to homogeneous features provided in an embodiment of the present invention;

[0036] Figure 3 A schematic diagram of the structure of the attention sub-network provided by an embodiment of the present invention;

[0037] Figure 4 A schematic diagram of the structure of a hierarchical feature extraction subnetwork provided by an embodiment of the present invention;

[0038] Figure 5 A schematic structural diagram of an indoor positioning device based on RSSI fingerprint is provided in an embodiment of the present invention. Specific implementation plan

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field based on the present invention are within the scope of protection of the present invention.

[0040] The wireless signals acquired by indoor fingerprint positioning systems can fluctuate due to changes in base station equipment. For example, the addition, removal, and relocation of base station equipment can lead to different compositions of fingerprint data in different time domains, exacerbating the occurrence of fingerprint mismatches and severely impacting the performance of indoor location services. Therefore, indoor positioning methods based on Wi-Fi fingerprints face the following challenges: the fingerprint feature spaces extracted from offline and online data are heterogeneous, resulting in insufficient robustness and discriminability, making it difficult to obtain accurate matching results.

[0041] In order to solve the above problems, the embodiments of the present invention provide an indoor positioning method and device based on RSSI fingerprint, which are described in detail below.

[0042] First, an indoor positioning method based on RSSI fingerprint provided by an embodiment of the present invention is described.

[0043] See also Figure 1 , which is a flow chart of an indoor fingerprint positioning method provided by an embodiment of the present invention. The method is applied to an electronic device with computing capabilities. Exemplarily, the method is applied to a computer. The method includes the following steps S101 to S104.

[0044] Step S101: Get the RSS data of the WIFI at the reference point with known indoor coordinates and the location to be located in different time domains, and set the coefficient K S Expand the fingerprint dimension and construct a homogeneous WIFI RSSI fingerprint feature space.

[0045] In the process of indoor fingerprint positioning, the RSSI values ​​of all received APs signals are obtained on all RPs, and the RSSI values ​​are used as different features to form WIFI RSSI fingerprints according to the different LAN addresses (Media Access Control Address, MAC). S The dimension is used to fill the feature values ​​of new APs in the online fingerprint, construct a homogeneous feature space in different time domains, and perform data normalization.

[0046] For example, if the total number of base stations APs that can be detected in an indoor positioning area is K, and the total number of base stations detected in the offline phase is K O , the number of offline fingerprints collected is n O The number of online fingerprints collected is n T If Figure 2 As shown, the i-th offline fingerprint is expressed as:

[0047]

[0048] Among them, when j>KO hour, remember The corresponding label is Source domain D S The composition is:

[0049]

[0050] For online fingerprints, j≤K O Partial Keep it unchanged and fill the RSS value of the new APs into the fingerprint reserved dimension, that is, j>K O Partial If K>K S , then delete j>K S Otherwise, zero is added after j>K. At this time, the i-th online fingerprint is expressed as:

[0051]

[0052] remember The corresponding label is Target domain D U The composition is:

[0053]

[0054] We record the RSS value of undetected APs as 0, and normalize the RSS values ​​of other APs and set the lower limit to 0.2. The normalized RSSI value of fingerprint No. j is:

[0055]

[0056] in, is the RSS value of AP j of the normalized n0 fingerprints and the average RSS value of AP j, and f(·) is the mapping function.

[0057] Step S102: Use the RSSI fingerprint at the reference point with known coordinates to train the attention network, obtain the attention matrix of the RSSI fingerprint of the location to be located in different time domains, reconstruct each RSSI fingerprint based on the attention matrix (Attention, A) coefficient, and increase the weight of important features.

[0058] Specifically, such as Figure 3 , build an attention sub-network, which first generates an attention matrix through self-attention to weight the RSSI fingerprint feature, and then randomly misses some parameter values ​​through the Dropout layer to simulate the absence of APs, and obtains the reconstructed RSSI fingerprint feature. The size of the reconstructed feature is the same as the input feature, which is K S dimension.

[0059] For example, the above normalized fingerprint features are passed through the attention layer, and the matrices Q, K, and V are actually: Q = w q *X+b q ,K=w k *X+b k ,V=w v *X+b v , where the trainable parameter w q 、w k 、w v 、b q 、b k 、b v For the similarity between Q and K, this paper uses the dot product method to calculate the A matrix, which is specifically expressed as:

[0060]

[0061] Then, through the Dropout layer, the new i-th source domain fingerprint and target domain fingerprint It can be expressed as:

[0062]

[0063] in are the i-th source domain fingerprints The corresponding attention matrix and the i-th target domain fingerprint The corresponding attention matrix, Drop(·) is processed by the Dropout layer.

[0064] Step S103: Using the reconstructed RSSI fingerprint at the reference point with known coordinates to train a hierarchical feature extraction network, extract the hierarchical features of the RSSI fingerprint of the location to be located in different time domains, and fuse them with the reconstructed features to improve the position discrimination of the fingerprint.

[0065] Specifically, a hierarchical feature extraction subnetwork is built, which extracts low-dimensional hierarchical features from the reconstructed RSSI fingerprint through a dimensionality-reduced FNN, and linearly fuses the extracted hierarchical features with the reconstructed features to obtain fused features with high position discrimination.

[0066] For example, the above reconstruction features The hierarchical features of the output are obtained by dimensionality reduction FNN, such as Figure 4 , its size is related to the extracted hierarchical information. For example, if the positioning area is an m-story building, the hierarchical feature dimension is m. Then the extracted hierarchical features are concatenated with the Ks-dimensional reconstructed features to obtain K S +m-dimensional fusion features.

[0067] Step S104: using the fused fingerprint at the reference point with known coordinates to train the positioning network, and obtaining position estimates of the fused fingerprint at the location to be positioned in different time domains.

[0068] Specifically, the RSSI data and the reference point coordinates obtained from the reference point are input into the indoor fingerprint positioning model together, and the pre-trained model updates and adjusts the model parameters; after obtaining the above-mentioned pre-trained model, the RSSI data obtained at the unknown coordinates in the same indoor scene is input to obtain the position information of the fingerprint feature to be located output by the fingerprint positioning model.

[0069] Exemplarily, in the offline stage, the K S +m-dimensional fusion fingerprint input positioning network, such as Figure 4 First, a dimension-enhanced FNN1 is used to embed the fingerprint feature into an n>2K S In the +2m+2 dimensional space, the fingerprint position is estimated by a dimension reduction FNN2, and the network is trained to map the fingerprint to the position. The overall loss is:

[0070] l=CE(Y f ,F)+MSE(Y c ,C),

[0071] Where CE is the cross entropy loss, MSE is the mean square error loss, F and C represent the floor and location coordinates respectively.

[0072]

[0073]

[0074] Among them, i and are the true labels and prediction results of the features to be located, respectively, i is the network input, and f(·) represents the positioning network map. By optimizing l, we can improve floor classification accuracy and positioning precision when APs vary. In the application phase, the collected RSSI data is input into the positioning model described above, and the model outputs the location prediction of the fingerprint feature to be located.

[0075] Corresponding to the aforementioned indoor fingerprint positioning method, an embodiment of the present invention further provides an indoor fingerprint positioning device.

[0076] See also Figure 5 , is a schematic diagram of the structure of an indoor positioning device based on RSSI fingerprint provided by an embodiment of the present invention, the device is applied to an electronic device with computing capabilities, and the device includes:

[0077] Acquisition module 501: used to obtain the RSS data of the WIFI at the reference point with known indoor coordinates and the location to be located in different time domains, and set the coefficient K S Expand the fingerprint dimension and construct a homogeneous WIFI RSSI fingerprint feature space.

[0078] Feature reconstruction module 502 based on attention mechanism: Use the RSSI fingerprint at the reference point with known coordinates to train the attention network, obtain the attention matrix of the RSSI fingerprint of the location to be located in different time domains, reconstruct each RSSI fingerprint based on the attention matrix coefficient, and increase the weight of important features.

[0079] Hierarchical feature extraction module 503: uses the reconstructed RSSI fingerprint at the reference point with known coordinates to train the hierarchical feature extraction network, extracts the hierarchical features of the RSSI fingerprint of the location to be located in different time domains, and fuses them with the reconstructed features to improve the position discrimination of the fingerprint.

[0080] Position prediction module 504: uses the fused fingerprint at the reference point with known coordinates to train the positioning network to obtain the position estimation of the fused fingerprint at the location to be positioned in different time domains.

[0081] In one embodiment of the present invention, the acquisition module 501 is specifically configured to:

[0082] The RSS data of the WIFI base station equipment at the reference point with known coordinates and the location to be located in different time domains are obtained by mobile phone, the RSS data is normalized to obtain RSSI data, and then the coefficient K is used to calculate the RSSI data. S Expand the fingerprint dimension and obtain a homogeneous WIFI RSSI fingerprint feature space.

[0083] In one embodiment of the present invention, the attention mechanism-based feature reconstruction module 502 is specifically configured to:

[0084] Build an attention sub-network, which first calculates the attention matrix parameters w,b. Then calculate the attention matrix A based on the RSSI fingerprint. S ,A T , based on the attention matrix A S ,A T Reconstruct each RSSI fingerprint. Finally, introduce the Dropout layer to increase the robustness of the reconstructed fingerprint to the disappearance of APs.

[0085] In one embodiment of the present invention, the hierarchical feature extraction module 503 is specifically configured to:

[0086] A hierarchical feature extraction subnetwork is built. The subnetwork first uses the attention subnetwork to reconstruct the RSSI fingerprint features, then uses the FNN to extract low-dimensional hierarchical features, and finally fuses the low-dimensional hierarchical features with the high-dimensional reconstructed features to obtain a fused fingerprint.

[0087] In one embodiment of the present invention, the location prediction module 504 is specifically configured to:

[0088] A positioning network is built. The fused fingerprint obtained by splicing the attention subnetwork with the hierarchical feature subnetwork is first passed through a dimensionality-increasing FNN and then a dimensionality-reducing FNN to obtain a position estimate. Joint training is then performed, and the fused fingerprint at a reference point with known coordinates is input to train the positioning network. The position estimate of the fused fingerprint at the target location is obtained in different time domains.

Claims

1. An indoor positioning method based on RSSI fingerprint, characterized in that: The method comprises: Get the RSS data of WIFI at the reference point with known indoor coordinates and the location to be located in different time domains, and set the coefficient K S Expand the fingerprint dimension and construct a homogeneous WIFI RSSI fingerprint feature space; The attention network is trained using the RSSI fingerprint at a reference point with known coordinates to obtain the attention matrix of the RSSI fingerprint at the location to be located in different time domains. Each RSSI fingerprint is reconstructed based on the attention matrix coefficient to increase the weight of important features. The hierarchical feature extraction network is trained using the reconstructed RSSI fingerprint at the reference point with known coordinates. The hierarchical features of the RSSI fingerprint of the location to be located in different time domains are extracted and fused with the reconstructed features to improve the position discrimination of the fingerprint. The positioning network is trained using the fused fingerprint at the reference point with known coordinates to obtain the position estimation of the fused fingerprint at the location to be positioned in different time domains.

2. The method according to claim 1, characterized in that The RSS data of the WIFI at the reference point with known indoor coordinates and the location to be located in different time domains is obtained, and the coefficient K is set. S Expand the fingerprint dimension and construct a homogeneous Wi-Fi RSSI fingerprint feature space, including: The RSS data of the WIFI base station equipment at the reference point with known coordinates and the location to be located in different time domains are obtained by mobile phone, the RSS data is normalized to obtain the RSSI data, and then the coefficient K is used to calculate the RSSI data. S Expand the fingerprint dimension and obtain a homogeneous WIFI RSSI fingerprint feature space.

3. The method according to claim 1, characterized in that The method uses the RSSI fingerprint at the reference point with known coordinates to train the attention network, obtains the attention matrix of the RSSI fingerprint of the location to be located in different time domains, reconstructs each RSSI fingerprint based on the attention matrix coefficient, and increases the weight of important features, including: Build an attention sub-network, which first calculates the attention matrix parameters w, b; then calculates the attention matrix A based on the RSSI fingerprint S , A T , based on the attention matrix A S , A T Reconstruct each RSSI fingerprint; finally, introduce the Dropout layer to increase the robustness of the reconstructed fingerprint to the disappearance of APs.

4. The method according to claim 1, wherein The method uses the reconstructed RSSI fingerprint at the reference point with known coordinates to train a hierarchical feature extraction network, extracts the hierarchical features of the RSSI fingerprint at the location to be located in different time domains, and fuses them with the reconstructed features to improve the position discrimination of the fingerprint, including: A hierarchical feature extraction subnetwork is built. The subnetwork first uses the attention subnetwork to reconstruct the RSSI fingerprint features, then uses the FNN to extract low-dimensional hierarchical features, and finally fuses the low-dimensional hierarchical features with the high-dimensional reconstructed features to obtain a fused fingerprint.

5. The method according to claim 1, wherein The method of training a positioning network using the fused fingerprint at a reference point with known coordinates to obtain position estimates of the fused fingerprint at the location to be positioned in different time domains includes: A positioning network is built, and the fused fingerprint obtained by splicing the attention sub-network and the hierarchical feature sub-network is first passed through a dimensionality-increasing FNN and then a dimensionality-reducing FNN to obtain a position estimate; and joint training is performed, and the fused fingerprint at a reference point with known coordinates is input to train the positioning network to obtain the position estimate of the fused fingerprint at the location to be located in different time domains.

6. An indoor positioning device based on RSSI fingerprint, characterized in that: The device comprises: Acquisition module: used to obtain the RSS data of WIFI at the reference point with known indoor coordinates and the location to be located in different time domains, and set the coefficient K S Expand the fingerprint dimension and construct a homogeneous WIFI RSSI fingerprint feature space; Feature reconstruction module based on attention mechanism: The attention network is trained using the RSSI fingerprint at the reference point with known coordinates, and the attention matrix of the RSSI fingerprint of the location to be located in different time domains is obtained. The RSSI fingerprint is reconstructed based on the attention matrix coefficient to increase the weight of important features. Hierarchical feature extraction module: uses the reconstructed RSSI fingerprint at the reference point with known coordinates to train the hierarchical feature extraction network, extracts the hierarchical features of the RSSI fingerprint of the location to be located in different time domains, and fuses them with the reconstructed features to improve the position discrimination of the fingerprint; Position prediction module: Use the fused fingerprint at the reference point with known coordinates to train the positioning network to obtain the position estimate of the fused fingerprint at the location to be positioned in different time domains.

7. The device according to claim 6, characterized in that The acquisition module is specifically used for: The RSS data of the WIFI base station equipment at the reference point with known coordinates and the location to be located in different time domains are obtained by mobile phone, the RSS data is normalized to obtain RSSI data, and then the coefficient K is used to calculate the RSSI data. S Expand the fingerprint dimension and obtain a homogeneous WIFI RSSI fingerprint feature space.

8. The device according to claim 6, characterized in that The feature reconstruction module based on the attention mechanism is specifically used for: Build an attention sub-network, which first calculates the attention matrix parameters w, b; then calculates the attention matrix A based on the RSSI fingerprint S , A T , based on the attention matrix A S , A T Reconstruct each RSSI fingerprint; finally, introduce the Dropout layer to increase the robustness of the reconstructed fingerprint to the disappearance of APs.

9. The device according to claim 6, characterized in that The hierarchical feature extraction module is specifically used for: A hierarchical feature extraction subnetwork is built. The subnetwork first uses the attention subnetwork to reconstruct the RSSI fingerprint features, then uses the FNN to extract low-dimensional hierarchical features, and finally fuses the low-dimensional hierarchical features with the high-dimensional reconstructed features to obtain a fused fingerprint.

10. The device according to claim 6, characterized in that The position prediction module is specifically used for: A positioning network is built, and the fused fingerprint of a reference point with known coordinates is input. The position estimate is obtained by first passing it through an up-dimensional FNN and then a down-dimensional FNN, and joint training is performed. After obtaining the above training model, the fused fingerprint of the location to be located in different time domains is input to obtain the position estimate of the fingerprint to be located.

Images