Indoor positioning method and device based on kernel function feature extraction and lasso algorithm

Through kernel function feature extraction and lasso algorithm, combined with KPCA and LASSO algorithm, the problem of low indoor positioning accuracy is solved, high-precision indoor positioning is achieved, and outdoor positioning is provided.

CN114866971BActive Publication Date: 2024-11-05CHINA UNIV OF PETROLEUM (EAST CHINA)
View PDF 14 Cites 0 Cited by

Patent Information

Application Number
CN202210497711.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-05-06
Publication Date
2024-11-05
Estimated Expiration
2042-05-06

AI Technical Summary

Technical Problem

The existing indoor positioning methods of fingerprint architecture type are not high enough due to the complexity of the indoor environment, multipath, shadow effect, personnel movement and channel congestion.

Method used

The KPCA method and lasso algorithm based on kernel function feature extraction are adopted. By obtaining offline feature fingerprint space and online feature position fingerprint space, the LASSO algorithm is used for regression analysis to improve positioning accuracy.

Benefits of technology

It improves the accuracy of indoor positioning, can better handle the properties of nonlinear data structures, improves the feature extraction ability and the accuracy of regression curves, provides a high-precision indoor positioning method, and supports outdoor positioning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114866971B_ABST
    Figure CN114866971B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of radio technology, and in particular to an indoor positioning method and device based on kernel function feature extraction and lasso algorithm, aiming to solve the problem that the positioning accuracy of existing fingerprint architecture type positioning methods is not high enough. The indoor positioning method and device based on kernel function feature extraction and lasso algorithm effectively extracts nonlinear features in data through KPCA method, estimates linear models with sparse parameters through LASSO algorithm, and reduces the variance of the model. The KPCA method and LASSO algorithm cooperate with each other to improve the feature extraction ability of traditional feature extraction methods and improve the accuracy of regression curves, and can better handle the inherent nonlinear data structure properties of sample space, and realize cluster analysis and regression analysis by nonlinearly mapping linear inseparable patterns in feature space to higher dimensional feature space, thereby further improving the accuracy of indoor positioning, and this high-precision indoor positioning method provides favorable support for outdoor positioning.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of radio technology, and in particular to an indoor positioning method and device based on kernel function feature extraction and lasso algorithm. Background Art

[0002] The emergence of Internet of Things technology has promoted a new round of information technology development, and wireless communication technology has also been continuously innovated. With the strong support of the country, multi-network integration technology based on WLAN, RFID, ZigBee, Bluetooth, mobile Internet and other technologies, D2D (Devices to Devices) technology has become the main development trend in the field of communication networks, so the network types are also becoming increasingly diversified, and the diversification of network types has also driven the diversification of network applications. In the early days of the Internet, the services that the network could provide were mainly text-based services. With the development of the network, Internet-based applications have developed from chat and e-mail to today's wide variety of application types, such as e-commerce services, multimedia services, etc. Traditional service models in various fields have also gradually realized online application forms through the "Internet +" platform. Among these many applications based on the network platform, most of them are inseparable from the user's location information, so the importance of location awareness technology is beyond doubt.

[0003] Location awareness technology can be divided into outdoor positioning technology and indoor positioning technology according to the different regions to be positioned. Traditional satellite-based positioning technologies, such as the domestic Beidou Satellite Navigation System (BDS), the US Global Positioning System (GPS), and the EU "Galileo" system, can well meet the needs of outdoor positioning, but when used for indoor positioning, the positioning effect is very poor due to the influence of indoor buildings on satellite signals. On the contrary, short-range wireless communication technology is more suitable for indoor positioning. Among them, Wireless Local Area Networks (WLAN) technology is often used for indoor positioning due to its characteristics of simple indoor installation, wide deployment, and low price.

[0004] In recent years, indoor positioning based on WLAN technology has become a research hotspot, among which the fingerprint architecture type positioning method has gradually become the mainstream of research. This fingerprint architecture type indoor positioning method uses the characteristic parameters of radio wave signals for positioning. However, due to the complexity of the indoor environment, the influence of factors such as multipath, shadow effect, movement of people, channel congestion and limited communication radius of nodes, wireless signals are easily attenuated and present time-varying characteristics, making the positioning accuracy of existing fingerprint architecture type positioning methods not high enough. Summary of the invention

[0005] The present invention provides an indoor positioning method and device based on kernel function feature extraction and lasso algorithm, so as to solve the problem that the positioning accuracy of the existing fingerprint architecture type positioning method is not high enough.

[0006] In order to achieve the above object, in a first aspect, the present invention provides an indoor positioning method based on kernel function feature extraction and lasso algorithm, the method comprising:

[0007] Get the original position fingerprint space F = {F 1 , F 2 , F 3 … F N} T , where F i = {rss i1 , rss i2 , rss i3 …rss in} T , i∈(1,N), N represents the number of reference nodes in the selected area, and the physical location of each reference node is l i (x i ,y i ), the physical location space of N reference nodes is L = (l 1 , l 2 , l 3 ,…l N ) T ; n represents the number of non-line-of-sight accessible points collected on each reference node; rss in represents the mean received signal strength (RSS mean) after sampling the nth accessible point of the i-th reference node p times;

[0008] The offline feature fingerprint space F′ is obtained by the kernel function feature extraction method (KPCA method), where Where K represents the kernel matrix; represents the modified kernel matrix; α represents The eigenvectors of by Schmidt orthogonalization;

[0009] Find some fingerprint points in the original position fingerprint space as landmark points to obtain the landmark point database S;

[0010] The new location fingerprint collected by the user is added to the landmark database S to form an online location fingerprint database O. The online location fingerprint database O obtains the online feature location fingerprint space O′ through the kernel function feature extraction method (KPCA method);

[0011] The LASSO algorithm is used to perform regression analysis on the offline feature fingerprint space F′, and two regression curves of the physical position coordinates x and y of the offline feature fingerprint space F′ are obtained with respect to the received signal strength (RSS). The Euler distance is compared between each row vector in the online feature position fingerprint space O′ and each row vector in the offline feature fingerprint space F′, and the position information in L corresponding to each row vector in the online feature position fingerprint space O′ and its most similar row vector in F′ is found;

[0012] After eliminating the known landmark coordinates in the online feature position fingerprint space O′, only one coordinate of the point to be measured remains, which is the real-time coordinate of the user.

[0013] In the above-mentioned indoor positioning method based on kernel function feature extraction and lasso algorithm, optionally, the offline feature fingerprint space F′ is obtained by the kernel function feature extraction method (KPCA method), specifically:

[0014] The original position fingerprint space F is used as the input space to calculate the kernel matrix K, where K ij = exp(|F i -F j | / -ε 2 ), K ij represents the element at the i-th row and j-th column in K, i, j <= N; ε represents the Gaussian kernel width;

[0015] By formula Calculate the modified kernel matrix Among them, l N Represents an N×N dimensional matrix, where the value of each element in the matrix is ​​1 / N;

[0016] Calculated by the function numpy.linalg.elg() The eigenvalues ​​and eigenvectors of The first k largest eigenvalues ​​of 1 ≥λ 2 …λ k-1 ≥λ k and the corresponding eigenvector v 1 ≥v 2 …v k-1 ≥ v k ;

[0017] By Schmidt orthogonalization, v 1 , v 2 … v k Orthogonalization of the unit gives α 1 , α 2 , …α k ;

[0018] By formula Calculate the offline feature fingerprint space F′, where α=(α 1 , α 2 , …α k ).

[0019] In the above-mentioned indoor positioning method based on kernel function feature extraction and lasso algorithm, optionally, some fingerprint points are found in the original position fingerprint space as landmark points to obtain the landmark point database S, specifically:

[0020] A Cartesian rectangular coordinate axis is established with the center of the selected area as the origin, and the received signal strength (RSS) on the coordinate axis is collected every one meter to obtain a landmark point database S.

[0021] In a second aspect, the present invention provides an indoor positioning device based on kernel function feature extraction and lasso algorithm, the device comprising:

[0022] The original position fingerprint space acquisition module is used to obtain the original position fingerprint space F = {F 1 , F 2 , F 3 … F N} T , where F i ={rss i1 , rss i2 , rss i3 … rss in} T , i∈(1,N), N represents the number of reference nodes in the selected area, and the physical location of each reference node is l i (x i ,y i ), the physical location space of N reference nodes is L = (l 1 , l 2 , l 3 ,… l N ) T ; n represents the number of non-line-of-sight accessible points collected on each reference node; rss in represents the mean received signal strength (RSS mean) after sampling the nth accessible point of the i-th reference node p times;

[0023] The offline feature fingerprint space acquisition module is used to obtain the offline feature fingerprint space F′ through the kernel function feature extraction method (KPCA method), where: Where K represents the kernel matrix; represents the modified kernel matrix; α represents The eigenvectors of by Schmidt orthogonalization;

[0024] A landmark point database acquisition module is used to find some fingerprint points in the original position fingerprint space as landmark points to obtain a landmark point database S;

[0025] The online feature position fingerprint space acquisition module is used to add the new position fingerprint collected by the user to the landmark database S to form an online position fingerprint database O. The online position fingerprint database O obtains the online feature position fingerprint space O′ through the kernel function feature extraction method (KPCA method);

[0026] The position information acquisition module is used to perform regression analysis on the offline feature fingerprint space F′ using the LASSO algorithm, obtain two regression curves of the physical position coordinates x and y of the offline feature fingerprint space F′ with respect to the received signal strength (RSS), compare the Euler distance between each row vector in the online feature position fingerprint space O′ and each row vector in the offline feature fingerprint space F′, and find the position information in L corresponding to each row vector in the online feature position fingerprint space O′ and its most similar row vector in F′;

[0027] The real-time coordinate acquisition module is used to remove the known landmark point coordinates in the online feature position fingerprint space O′ and leave only one coordinate of the point to be measured, that is, the real-time coordinate of the user.

[0028] In the above-mentioned indoor positioning device based on kernel function feature extraction and lasso algorithm, optionally, the offline feature fingerprint space F′ is obtained by the kernel function feature extraction method (KPCA method), specifically:

[0029] The original position fingerprint space F is used as the input space to calculate the kernel matrix K, where K ij = exp(|F i -F j | / -ε 2 ), K ij represents the element at the i-th row and j-th column in K, i, j <= N; ε represents the Gaussian kernel width;

[0030] By formula Calculate the modified kernel matrix Among them, l N Represents an N×N dimensional matrix, where the value of each element in the matrix is ​​1 / N;

[0031] Calculated by the function numpy.linalg.elg() The eigenvalues ​​and eigenvectors of The first k largest eigenvalues ​​of 1 ≥λ 2 … λ k-1 ≥λ kand the corresponding eigenvector v 1 ≥v 2 … v k-1 ≥ v k ;

[0032] By Schmidt orthogonalization, v 1 , v 2 … v k Orthogonalization of the unit gives α 1 , α 2 , … α k ;

[0033] By formula Calculate the offline feature fingerprint space F′, where α=(α 1 , α 2 , …α k ).

[0034] In the above-mentioned indoor positioning device based on kernel function feature extraction and lasso algorithm, optionally, some fingerprint points are found in the original position fingerprint space as landmark points to obtain the landmark point database S, specifically:

[0035] A Cartesian rectangular coordinate axis is established with the center of the selected area as the origin, and the received signal strength (RSS) on the coordinate axis is collected every one meter to obtain a landmark point database S.

[0036] The indoor positioning method and device based on kernel function feature extraction and lasso algorithm provided by the present invention effectively extract nonlinear features in data through the KPCA method, estimate the linear model with sparse parameters through the LASSO algorithm, and reduce the variance of the model. The KPCA method and the LASSO algorithm cooperate with each other to improve the feature extraction ability of the traditional feature extraction method and improve the accuracy of the regression curve. The KPCA method and the LASSO algorithm can better handle the inherent nonlinear data structure properties of the sample space, and the linear inseparable pattern of the feature space is nonlinearly mapped to a higher-dimensional feature space to realize cluster analysis and regression analysis, thereby further improving the accuracy of indoor positioning. The high-precision indoor positioning method provides favorable support for outdoor positioning.

[0037] The construction of the present invention and other objects and advantageous effects thereof will become more apparent through the description of the preferred embodiments in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A flowchart of an indoor positioning method based on kernel function feature extraction and lasso algorithm provided in an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of the structure of an indoor positioning device based on kernel function feature extraction and lasso algorithm provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme in the embodiment of the present invention will be described in more detail below in conjunction with the drawings in the preferred embodiments of the present invention. In the drawings, the same or similar reference numerals throughout represent the same or similar parts or parts with the same or similar functions. The described embodiments are part of the embodiments of the present invention, not all of them. The embodiments described below with reference to the drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limitations on the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The embodiments of the present invention are described in detail below in conjunction with the drawings.

[0042] like Figure 1 As shown, in a first aspect, the present invention provides an indoor positioning method based on kernel function feature extraction and lasso algorithm, comprising:

[0043] Step 110: Obtain the original position fingerprint space F = {F 1 , F 2 , F 3 … F N} T , where F i = {rss i1 , rss i2 , rss i3 … rss in} T , i∈(1,N), N represents the number of reference nodes in the selected area, and the physical location of each reference node is l i (x i ,y i ), the physical location space of N reference nodes is L = (l 1 , l2 , l 3 ,… l N ) T ; n represents the number of non-line-of-sight accessible points collected on each reference node; rss in represents the mean received signal strength (RSS mean) after sampling the nth accessible point of the i-th reference node p times;

[0044] It should be noted that the selected area includes N reference nodes, and the physical location of each reference node is l i (x i ,y i ), the physical positions of N reference nodes constitute the position space L = (l 1 , l 2 , l 3 ,… l N ) T ;

[0045] At each reference node, the received signal strength (RSS) and media access control address (MAC) information of n non-line-of-sight access points (AP) are collected. Here, MAC information is used as the identifier of each WiFi source. MAC information is only used to mark each collected WiFi source in the offline collection stage. In the actual algorithm, MAC information does not participate in any calculation;

[0046] Perform p samplings at each accessible point, and use the mean received signal strength (RSS mean) after p samplings as the reference node l i (x i ,y i ) is the original location fingerprint information. Such an original location fingerprint information is an n-dimensional vector F i ={rss i1 , rss i2 , rss i3 … rss in} T , i∈ (1, N), rss in represents the mean received signal strength (RSS mean) after sampling the nth accessible point of the i-th reference node p times; the original position fingerprint information of all reference nodes is stored in the relevant database to form an N×n dimensional matrix, that is, F = {F 1 , F 2 , F 3 … F N} T, each row vector in the matrix F represents the original position fingerprint vector of a reference node. We call this F the original position fingerprint space. Each row vector in the matrix F represents the original position fingerprint vector of a reference node. The dimension of each row vector is the number of accessible points that can be collected in this space. Each row vector corresponds to a position information in the physical position space L.

[0047] Step 120: Obtain offline feature fingerprint space F′ by kernel function feature extraction method (KPCA method), where: Where K represents the kernel matrix; represents the modified kernel matrix; α represents The eigenvectors of through Schmidt orthogonalization are:

[0048] The original position fingerprint space F is used as the input space to calculate the kernel matrix K, where K ij = exp(|F i -F j | / -ε 2 ), K ij represents the element at the i-th row and j-th column in K, i, j <= N; ε represents the Gaussian kernel width, and K is an N×N dimensional matrix;

[0049] By formula Calculate the modified kernel matrix Among them, l N represents an N×N dimensional matrix, l N It is a very special matrix. The value of each element in the matrix is ​​1 / N, which means that the value that determines l N The size of each element is its dimension;

[0050] Calculated by the function numpy.linalg.elg() The eigenvalues ​​and eigenvectors of The first k largest eigenvalues ​​of 1 ≥λ 2 … λ k-1 ≥λ k and the corresponding eigenvector v 1 ≥v 2 … v k-1 ≥ v k ;

[0051] By Schmidt orthogonalization, v 1 , v 2 … v k Orthogonalization of the unit gives α 1 , α 2 , … α k ;

[0052] By formula Calculate the offline feature fingerprint space F′, where α=(α 1 , α 2 , … α k ).

[0053] It should be noted that The eigenvalues ​​and eigenvectors are obtained through the built-in python function numpy.linalg.elg(); the offline feature fingerprint space F′ is an N×K dimensional matrix.

[0054] The advantage of using the KPCA method is that the KPCA method can transform linearly inseparable data into a high-dimensional space through nonlinear mapping. Generally, high-dimensional space is easier to be linearly separable. Then, standard PCA is used in the high-dimensional space to map it into another low-dimensional space so that it can be divided by a linear classifier.

[0055] Step 130: Find some fingerprint points in the original position fingerprint space as landmark points to obtain a landmark point database S. The landmark point selection rule is:

[0056] A Cartesian rectangular coordinate axis is established with the center of the selected area as the origin, and the received signal strength (RSS) on the coordinate axis is collected every one meter to obtain a landmark point database S.

[0057] It should be noted that the landmark database S is independent of the original position fingerprint space F. The landmark database S can also be established by directly selecting the position fingerprint vector of the reference node position on the coordinate axis in the original position fingerprint space F.

[0058] Step 140: adding the new location fingerprint collected by the user to the landmark database S to form an online location fingerprint database O, and obtaining an online feature location fingerprint space O′ from the online location fingerprint database O through a kernel function feature extraction method (KPCA method);

[0059] It should be noted that the APP in the user's hand will actively collect the location fingerprint information of the user's location. After the user collects the location fingerprint of each point, the fingerprint will be added to the landmark database S. We call this collected new fingerprint the positioning stage fingerprint. The positioning stage fingerprint and the landmark database S are combined into an online location fingerprint database O. The online location fingerprint database O obtains the online feature location fingerprint space O′ through the kernel function feature extraction method (KPCA method). The reason for combining the positioning stage fingerprint and the landmark database S is that the KPCA method cannot extract the feature value from a point, so it is necessary to add additional known information to meet the requirements of the KPCA method; the KPCA method used to obtain the online feature location fingerprint space O′ is the same as the KPCA method used to obtain the offline feature fingerprint space F′ in step 120.

[0060] Step 150: Use the LASSO algorithm to perform regression analysis on the offline feature fingerprint space F′, obtain two regression curves of the physical position coordinates x and y of the offline feature fingerprint space F′ with respect to the received signal strength (RSS), compare the Euler distance between each row vector in the online feature position fingerprint space O′ and each row vector in the offline feature fingerprint space F′, and find the position information in L corresponding to each row vector in the online feature position fingerprint space O′ and its most similar row vector in F′;

[0061] It should be noted that in the process of obtaining the offline feature fingerprint space F′ and the online feature position fingerprint space O′, the correspondence between F′, O′ and the physical position space L remains unchanged. The LASSO algorithm can be used to obtain the correspondence between the eigenvalues ​​in F′ and L. Then, the Euler distance between each row vector in the online feature position fingerprint space O′ and each row vector in the offline feature fingerprint space F′ is compared, and the row vector in F′ that is most similar to each row vector in the online feature position fingerprint space O′ is found. The position information in L corresponding to the most similar row vector in F′ is the positioning position we need.

[0062] The LASSO algorithm is a compression estimation. It obtains a more refined model by constructing a penalty function, which compresses some regression coefficients, that is, it forces the sum of the absolute values ​​of the coefficients to be less than a fixed value, and sets some regression coefficients to zero, thereby retaining the advantages of subset shrinkage. It is a biased estimation for processing data with complex collinearity. LASSO regression makes some coefficients smaller, and even some coefficients with smaller absolute values ​​directly become 0, so it is particularly suitable for the selection and reduction of the number of parameters, and is therefore used to estimate sparse parameters.

[0063] Step 160: After eliminating the known landmark point coordinates in the online feature position fingerprint space O′, only one coordinate of the point to be measured remains, namely, the real-time coordinate of the user.

[0064] It should be noted that by setting a timestamp to ensure that steps 140 to 160 are continuously cycled, real-time indoor positioning can be achieved; when the user moves in a specific space, the process will be actively triggered every second to ensure the real-time positioning.

[0065] The indoor positioning method and device based on kernel function feature extraction and lasso algorithm provided by the present invention effectively extract nonlinear features in data through the KPCA method, estimate the linear model with sparse parameters through the LASSO algorithm, and reduce the variance of the model. The KPCA method and the LASSO algorithm cooperate with each other to improve the feature extraction ability of the traditional feature extraction method and improve the accuracy of the regression curve. The KPCA method and the LASSO algorithm can better handle the inherent nonlinear data structure properties of the sample space, and the linear inseparable pattern of the feature space is nonlinearly mapped to a higher-dimensional feature space to realize cluster analysis and regression analysis, thereby further improving the accuracy of indoor positioning. The high-precision indoor positioning method provides favorable support for outdoor positioning.

[0066] like Figure 2 As shown, in a second aspect, the present invention provides an indoor positioning device based on kernel function feature extraction and lasso algorithm, comprising:

[0067] The original position fingerprint space acquisition module 201 is used to obtain the original position fingerprint space F = {F 1 , F 2 , F 3 …F N} T , where F i ={rss i1 , rss i2 , rss i3 … rss in} T , i∈(1,N), N represents the number of reference nodes in the selected area, and the physical location of each reference node is l i (x i ,y i ), the physical location space of N reference nodes is L = (l 1 , l 2 , l 3 ,…l N ) T ; n represents the number of non-line-of-sight accessible points collected on each reference node; rss inrepresents the mean received signal strength (RSS mean) after sampling the nth accessible point of the i-th reference node p times;

[0068] The offline feature fingerprint space acquisition module 202 is used to obtain the offline feature fingerprint space F′ by using the kernel function feature extraction method (KPCA method), wherein: Where K represents the kernel matrix; represents the modified kernel matrix; α represents The eigenvectors of through Schmidt orthogonalization are:

[0069] The original position fingerprint space F is used as the input space to calculate the kernel matrix K, where K ij = exp(|F i -F j | / -ε 2 ), K ij represents the element at the i-th row and j-th column in K, i, j <= N; ε represents the Gaussian kernel width, and K is an N×N dimensional matrix;

[0070] By formula Calculate the modified kernel matrix Among them, l N represents an N×N dimensional matrix, l N It is a very special matrix. The value of each element in the matrix is ​​1 / N, which means that the value that determines l N The size of each element is its dimension;

[0071] Calculated by the function numpy.linalg.elg() The eigenvalues ​​and eigenvectors of The first k largest eigenvalues ​​of 1 ≥λ 2 … λ k-1 ≥λ k and the corresponding eigenvector v 1 ≥v 2 … v k-1 ≥ v k ;

[0072] By Schmidt orthogonalization, v 1 , v 2 … v k Orthogonalization of the unit gives α 1 , α 2 , … α k ;

[0073] By formula Calculate the offline feature fingerprint space F′, where α=(α 1 , α2 , …α k ).

[0074] The landmark database acquisition module 203 is used to find some fingerprint points in the original position fingerprint space as landmark points to obtain a landmark database S. The landmark point selection rule is:

[0075] A Cartesian rectangular coordinate axis is established with the center of the selected area as the origin, and the received signal strength (RSS) on the coordinate axis is collected every one meter to obtain a landmark point database S.

[0076] The online feature position fingerprint space acquisition module 204 is used to add the new position fingerprint collected by the user to the landmark database S to form an online position fingerprint database O. The online position fingerprint database O obtains the online feature position fingerprint space O′ through the kernel function feature extraction method (KPCA method);

[0077] The position information acquisition module 205 is used to perform regression analysis on the offline feature fingerprint space F′ using the LASSO algorithm to obtain two regression curves of the physical position coordinates x and y of the offline feature fingerprint space F′ with respect to the received signal strength (RSS), compare the Euler distance between each row vector in the online feature position fingerprint space O′ and each row vector in the offline feature fingerprint space F′, and find the position information in L corresponding to each row vector in the online feature position fingerprint space O′ and the row vector in F′ that is most similar to it;

[0078] The real-time coordinate acquisition module 206 is used to remove the known marker point coordinates in the online feature position fingerprint space O′ and leave only one point coordinate to be measured, namely the real-time coordinate of the user.

[0079] It should be noted that the indoor positioning device based on kernel function feature extraction and lasso algorithm provided in the above embodiment and the indoor positioning method embodiment based on kernel function feature extraction and lasso algorithm belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An indoor positioning method based on kernel function feature extraction and lasso algorithm, characterized in that: The method comprises: Get the original position fingerprint space F = {F1, F2, F3...F N } T ,in, L=(l1,l2,,3,...l N ) T ; n represents the number of non-line-of-sight accessible points collected on each reference node; rss in represents the mean received signal strength (RSS mean) after sampling the nth accessible point of the i-th reference node p times; The offline feature fingerprint space F′ is obtained by the kernel function feature extraction method (KPCA method), where Where K represents the kernel matrix; represents the modified kernel matrix; α represents The eigenvectors of by Schmidt orthogonalization; Find some fingerprint points in the original position fingerprint space as landmark points to obtain the landmark point database S; The new location fingerprint collected by the user is added to the landmark database S to form an online location fingerprint database O. The online location fingerprint database O obtains the online feature location fingerprint space O′ through the kernel function feature extraction method (KPCA method); The LASSO algorithm is used to perform regression analysis on the offline feature fingerprint space F′, and two regression curves of the physical position coordinates x and y of the offline feature fingerprint space F′ are obtained with respect to the received signal strength (RSS). The Euler distance is compared between each row vector in the online feature position fingerprint space O′ and each row vector in the offline feature fingerprint space F′, and the position information in L corresponding to each row vector in the online feature position fingerprint space O′ and its most similar row vector in F′ is found; After removing the known landmark coordinates in the online feature position fingerprint space O′, only one coordinate of the point to be measured remains, namely the real-time coordinates of the user; the offline feature fingerprint space F′ is obtained by the kernel function feature extraction method (KPCA method), specifically: The original position fingerprint space F is used as the input space to calculate the kernel matrix K, where K ij =exp(|F i -F j | / -ε 2 ), K ij represents the element at the i-th row and j-th column in K, i, j <= N; ε represents the Gaussian kernel width; By formula Calculate the modified kernel matrix Among them, l N Represents an N×N dimensional matrix, where the value of each element in the matrix is ​​1 / N; Calculated by the function numpy.linalg.elg() The eigenvalues ​​and eigenvectors of The first k largest eigenvalues ​​of k-1 ≥λ k and the corresponding eigenvector v1≥v2…v k-1 ≥v k ; By Schmidt orthogonalization, v1, v2…v k Orthogonalization of the unit yields α1, α2, …α k ; By formula Calculate the offline feature fingerprint space F′, where α=(α1,α2,…α k ); The method of finding some fingerprint points in the original position fingerprint space as landmark points to obtain the landmark point database S is specifically as follows: A Cartesian rectangular coordinate axis is established with the center of the selected area as the origin, and the received signal strength (RSS) on the coordinate axis is collected every one meter to obtain a landmark point database S.

2. An indoor positioning device based on kernel function feature extraction and lasso algorithm, characterized in that: The device comprises: The original position fingerprint space acquisition module is used to obtain the original position fingerprint space The physical location space of the reference node is L = (l1, l2,,3,...l N ) T ; n represents the number of non-line-of-sight accessible points collected on each reference node; rss in represents the mean received signal strength (RSS mean) after sampling the nth accessible point of the i-th reference node p times; The offline feature fingerprint space acquisition module is used to obtain the offline feature fingerprint space F′ through the kernel function feature extraction method (KPCA method), where: Where K represents the kernel matrix; represents the modified kernel matrix; α represents The eigenvectors of by Schmidt orthogonalization; A landmark point database acquisition module is used to find some fingerprint points in the original position fingerprint space as landmark points to obtain a landmark point database S; The online feature position fingerprint space acquisition module is used to add the new position fingerprint collected by the user to the landmark database S to form an online position fingerprint database O. The online position fingerprint database O obtains the online feature position fingerprint space O′ through the kernel function feature extraction method (KPCA method); The position information acquisition module is used to perform regression analysis on the offline feature fingerprint space F′ using the LASSO algorithm, obtain two regression curves of the physical position coordinates x and y of the offline feature fingerprint space F′ with respect to the received signal strength (RSS), compare the Euler distance between each row vector in the online feature position fingerprint space O′ and each row vector in the offline feature fingerprint space F′, and find the position information in L corresponding to each row vector in the online feature position fingerprint space O′ and its most similar row vector in F′; The real-time coordinate acquisition module is used to remove the known landmark coordinates in the online feature position fingerprint space O′ and leave only one coordinate of the point to be measured, that is, the real-time coordinate of the user; The offline feature fingerprint space F′ is obtained by the kernel function feature extraction method (KPCA method), specifically: The original position fingerprint space F is used as the input space to calculate the kernel matrix K, where K ij =exp(|F i -F j | / -ε 2 ), K ij represents the element at the i-th row and j-th column in K, i, j <= N; ε represents the Gaussian kernel width; By formula Calculate the modified kernel matrix Among them, l N Represents an N×N dimensional matrix, where the value of each element in the matrix is ​​1 / N; Calculated by the function numpy.linalg.elg() The eigenvalues ​​and eigenvectors of The first k largest eigenvalues ​​of k-1 ≥λ k and the corresponding eigenvector v1≥v2…v k-1,1 ≥v k ; By Schmidt orthogonalization, v1, v2, ... k Orthogonalization of the unit yields α1, α2, …α k ; By formula Calculate the offline feature fingerprint space F′, where α=(α1,α2,…α k ); The method of finding some fingerprint points in the original position fingerprint space as landmark points to obtain the landmark point database S is specifically as follows: A Cartesian rectangular coordinate axis is established with the center of the selected area as the origin, and the received signal strength (RSS) on the coordinate axis is collected every one meter to obtain a landmark point database S.

Citation Information

Patent Citations

  • SVM-KNN (Support Vector Machine-K Nearest Neighbor)-based indoor positioning method

    CN104619014A

  • WIFI indoor weighted K nearest neighbor positioning algorithm based on kernel function main feature extraction

    CN105657823A

  • Terminal positioning method, device and system

    CN106550447A

  • Indoor positioning method based on dual-frequency Wi-Fi signal

    CN107347181A

  • Indoor fingerprint positioning method based on regional partition and sparse support vector regression

    CN108717175A