CN103079269A - LDE (Linear Discriminant Analysis) algorithm-based WiFi (Wireless Fidelity) indoor locating method - Google Patents

Abstract

A WiFi indoor positioning method based on an LDE algorithm relates to an indoor positioning method. It aims to solve the problem of poor real-time positioning of existing WiFi indoor positioning methods. The implementation process of this method is divided into two stages: offline stage: the construction of the WiFi network, the measurement of RSS, and the construction of the Radio Map; the intrinsic dimension estimation method is used to estimate the intrinsic dimension of the Radio Map; the LDE algorithm is used to estimate the radio map Dimensionality reduction processing, obtain the Radio Map and feature transformation matrix after dimensionality reduction, obtain the optimal dimensionality reduction result and the corresponding feature transformation matrix as the matching database and the corresponding RSS transformation matrix in the online stage. Online stage: Perform feature transformation on the RSS received at the test point, and use the KNN algorithm to match with the reduced-dimensional Radio Map to obtain the predicted coordinates of the test point. The invention is suitable for indoor positioning.

Description

WiFi indoor orientation method based on the LDE algorithm

Technical field

The present invention relates to a kind of indoor orientation method.

Background technology

Along with WLAN is popularized the extensive of worldwide develop rapidly and mobile terminal device, many indoor positioning relevant technology and application have appearred in recent years.Because the complexity of multipath effect, signal attenuation and indoor positioning environment, be difficult to reach high-precision indoor positioning requirement based on the indoor orientation method of traditional signal propagation model.Based on the time of advent (Time of Arrival), the time of advent poor (Time Difference of Arrival) or arrive angle (Angles of Arrival) although etc. localization method can substantially satisfy the positioning accuracy demand, yet all need locating terminal that extra hardware device support is arranged, have larger limitation, thereby cause not popularized based on the indoor locating system of above-mentioned a few class localization methods.

At present, the WiFi indoor orientation method based on WLAN location fingerprint (Finger Print) is widely applied.The network establishing method of the method is with low cost, and it uses 2.4GHz ISM (Industrial Science Medicine) common frequency band and need not to add the location survey specialized hardware on existing utility.Access point (the Access Point that only need to receive by wireless network card and the corresponding software measurement of portable terminal, AP) signal strength signal intensity (Received Signal Strength, RSS), make up thus network signal coverage diagram (Radio Map), and then predict the coordinate of mobile subscriber present position by matching algorithm, or relative position.

Yet include huge data message by the Radio Map that this mode is set up, and along with locating area enlarges, Radio Map may (select) to be the index situation and increase according to position matching mode and algorithm.Obtain related data characteristic information as much as possible and for whole system, can promote positioning accuracy, but process a large amount of characteristic informations and increase the algorithm expense, location algorithm can't effectively operation on the limited portable terminal of disposal ability, simultaneously some characteristic information may be not act on even negative effect arranged for the location, cause matching efficiency to reduce, thereby cause mating the realization more complex of location algorithm, and positioning accuracy descends.

When the number of AP increases and the reference point (Reference Point) of location when increasing, the data message of Radio Map increases.The information of the AP number that represents among the Radio Map at this moment, has represented the dimension of data.Therefore, when the increase of AP number, Radio Map has just become high dimensional data.For alleviating the burden of processing high dimensional data, dimension-reduction algorithm is one of effective solution.High dimensional data may comprise a lot of features, and these features are all being described same things, and these features are closely to link to each other to a certain extent.As when from all angles same object being taken pictures simultaneously, the data that obtain just contain overlapping information.If can obtain nonoverlapping expression of some simplification of these data, will greatly improve data and process the efficient of operation and improve to a certain extent accuracy.The purpose of dimension-reduction algorithm also is to improve the treatment effeciency of high dimensional data just.

Except can reduced data can efficiently processing, dimension reduction method can also be realized data visualization.Because a lot of statistical very poor for the accuracy of optimal solution with machine learning algorithm, the visual application of dimensionality reduction can make the user can actually see the space structure of high dimensional data and the ability of algorithm output, has very strong using value.

A lot of dimension-reduction algorithms based on different purposes are arranged at present, include linearity and nonlinear reductive dimension algorithm.Wherein PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis) are typical linear dimension-reduction algorithms.This class algorithm has good result for the high dimensional data with linear structure, but does not have good result for the high dimensional data of nonlinear organization.Typical nonlinear reductive dimension algorithm is with manifold learning (Manifold Learning) algorithm.3 pieces have been delivered relevant for the manifold learning arithmetic that has proposed 2 kinds of classics in the manifold learning arithmetic: LLE (Local Linear Embedding) and ISOMAP (Isometric Mapping) with first phase on the Science magazines in 2000.Thus, various manifold learning arithmetic based on different criterions be suggested and some manifold learning arithmetic be applied to the image processing method face.

Large, the online positioning stage computation complexity of the Radio Map database that present WiFi indoor orientation method exists is high-leveled and difficult the problem such as to realize, the real-time of location is poor at portable terminal.

Summary of the invention

The present invention is for the poor problem of the real-time of the location that solves existing WiFi indoor orientation method, thereby a kind of WiFi indoor orientation method based on the LDE algorithm is provided.

Based on the WiFi indoor orientation method of LDE algorithm, it is realized by following steps:

Step 1, arrange N access point AP for indoor environment, guarantee that any point is covered by the signal that one or more access point AP sends in the described environment, described N access point AP composition WiFi network; N evenly is set in described indoor environment _RPIndividual reference point; N and N _RPBe positive integer;

Step 2, to choose a reference point be that the origin of coordinates is set up the two-dimensional direct angle coordinate system, obtains N _RPThe coordinate position of individual reference point in this two-dimensional direct angle coordinate system utilizes the signal receiver collection from the signal strength signal intensity RSS value of each access point AP in each reference point in off-line phase, and as the position feature information of this access point AP; And according to the position feature information architecture indoor coverage of signal figure Radio Map of N access point AP;

Step 3, employing intrinsic dimension algorithm for estimating carry out the intrinsic dimensional analysis to the indoor coverage of signal figure Radio Map that step 2 obtains, and obtain intrinsic dimensional analysis result;

Step 4, the intrinsic dimensional analysis result who obtains according to step 3 adopt the LED algorithm with all the reference point dimensionality reductions in the indoor coverage of signal figure Radio Map to the intrinsic dimension, obtain the eigentransformation matrix, and generate the Radio map Radio Map behind the dimensionality reduction ^*

Step 5, at on-line stage, measure the signal strength signal intensity RSS value of wanting anchor point in the indoor environment, and with the eigentransformation matrix multiple of this signal strength signal intensity RSS value and step 4 acquisition, picked up signal intensity transformation value RSS ^*

Step 6, the signal strength signal intensity transformed value RSS that adopts the KNN algorithm that step 5 is obtained ^*Radio map Radio Map behind the dimensionality reduction that generates with step 4 ^*Carry out location matches, obtain to want the position coordinates of anchor point, finish the indoor positioning of wanting anchor point.

Adopting intrinsic dimension algorithm for estimating in the step 3 is the characteristic value estimation technique.

Adopt intrinsic dimension algorithm for estimating to be the bag number estimation technique in the step 3.

Adopting intrinsic dimension algorithm for estimating in the step 3 is the geodesic curve minimal spanning tree algorithm.

It is to pass through formula that the indoor coverage of signal figure Radio Map that adopts the geodesic curve minimal spanning tree algorithm that step 2 is obtained carries out the intrinsic dimensional analysis:

$d_{\mathrm{intrinsicdim}}=\frac{1}{1-a}$

Realize; In the formula: d _IntrinsicdimBe intrinsic dimensional analysis result; A represents the slope of the linear fit expression formula of minimum spanning tree.

Radio map Radio Map after obtaining the eigentransformation matrix in the step 4 and generating dimensionality reduction ^*Between the pass be:

Radio Map ^*＝V′·Radio Map。

The signal strength signal intensity transformed value RSS that adopts the KNN algorithm that step 5 is obtained ^*Radio map Radio Map behind the dimensionality reduction that generates with step 4 ^*The method of carrying out location matches is to pass through formula:

$(x^{\′},y^{\′})=\frac{1}{K}\underset{i=1}{\overset{K}{\mathrm{\Σ}}}(x_i,y_i)$

Realize;

In the formula: (x ', y ') for wanting the coordinate of anchor point, (x _i, y _i) be the coordinate of i Neighbor Points, i is positive integer; K is the sum of Neighbor Points in the KNN algorithm.

WiFi indoor positioning real-time of the present invention is high.Simultaneously, the present invention adopt the LDE algorithm with Radio Map dimensionality reduction to the intrinsic dimension, it is large to have reduced the Radio Map data volume that exists in the existing WiFi indoor orientation method, and has reduced online positioning stage computation complexity, makes it be easy to realize at portable terminal.

Description of drawings

Fig. 1 is the experiment scene schematic diagram described in the embodiment one.Fig. 2 is the signal flow schematic diagram of the inventive method.

Embodiment

Embodiment one, in conjunction with Fig. 2 this embodiment is described, based on the WiFi indoor orientation method of LDE algorithm, it is realized by following steps:

Step 1, arrange N access point AP for indoor environment, guarantee that any point is covered by the signal that one or more access point AP sends in the described environment, described N access point AP composition WiFi network; N evenly is set in described indoor environment _RPIndividual reference point; N and N _RPBe positive integer;

Step 2, to choose a reference point be that the origin of coordinates is set up the two-dimensional direct angle coordinate system, obtains N _RPThe coordinate position of individual reference point in this two-dimensional direct angle coordinate system utilizes the signal receiver collection from the signal strength signal intensity RSS value of each access point AP in each reference point in off-line phase, and as the position feature information of this access point AP; And according to the position feature information architecture indoor coverage of signal figure Radio Map of N access point AP;

Step 3, employing intrinsic dimension algorithm for estimating carry out the intrinsic dimensional analysis to the indoor coverage of signal figure Radio Map that step 2 obtains, and obtain intrinsic dimensional analysis result;

Step 4, the intrinsic dimensional analysis result who obtains according to step 3 adopt the LED algorithm with all the reference point dimensionality reductions in the indoor coverage of signal figure Radio Map to the intrinsic dimension, obtain the eigentransformation matrix, and generate the Radio map Radio Map behind the dimensionality reduction ^*

Step 5, at on-line stage, measure the signal strength signal intensity RSS value of wanting anchor point in the indoor environment, and with the eigentransformation matrix multiple of this signal strength signal intensity RSS value and step 4 acquisition, picked up signal intensity transformation value RSS ^*

Step 6, the signal strength signal intensity transformed value RSS that adopts the KNN algorithm that step 5 is obtained ^*Radio map Radio Map behind the dimensionality reduction that generates with step 4 ^*Carry out location matches, obtain to want the position coordinates of anchor point, finish the indoor positioning of wanting anchor point.

Adopting intrinsic dimension algorithm for estimating in the step 3 is the characteristic value estimation technique, the bag number estimation technique or geodesic curve minimal spanning tree algorithm.

It is to pass through formula that the indoor coverage of signal figure Radio Map that adopts the geodesic curve minimal spanning tree algorithm that step 2 is obtained carries out the intrinsic dimensional analysis:

$d_{\mathrm{intrinsicdim}}=\frac{1}{1-a}$

Realize; In the formula: d _IntrinsicdimBe intrinsic dimensional analysis result; A represents the slope of the linear fit expression formula of minimum spanning tree.

Obtain in the step 4 eigentransformation matrix V ' with the generation dimensionality reduction after Radio map Radio Map ^*Between the pass be:

Radio Map ^*＝V′·Radio Map。

The signal strength signal intensity transformed value RSS that adopts the KNN algorithm that step 5 is obtained ^*Radio map Radio Map behind the dimensionality reduction that generates with step 4 ^*The method of carrying out location matches is to pass through formula:

$(x^{\′},y^{\′})=\frac{1}{K}\underset{i=1}{\overset{K}{\mathrm{\Σ}}}(x_i,y_i)$

Realize;

In the formula: (x ', y ') for wanting the coordinate of anchor point, (x _i, y _i) be the coordinate of i Neighbor Points, i is positive integer; K is the sum of Neighbor Points in the KNN algorithm.

In the present embodiment, obtaining by following step of the intrinsic dimension of Radio Map realizes:

The intrinsic dimension is the number of carrying out the independent variable of eigenspace dimension and the required minimum of space reconstruction for high dimensional data.In concrete Practical Calculation, because the intrinsic of high dimensional data and not obvious is not to seek to obtain definite intrinsic dimension usually, but seeks to estimate the credible value of intrinsic dimension.Specifically, a given sample from higher dimensional space, the central task of intrinsic dimension algorithm for estimating and important content are exactly the intrinsic dimension of determining this higher-dimension structure by these sample datas.

The estimation of the intrinsic dimension of Radio Map is the important input parameter of LDE algorithm, and whether this result who is related to dimensionality reduction can represent the feature of the higher dimensional space of Radio Map, and therefore the estimation of intrinsic dimension is most important accurately and effectively.At present, intrinsic dimension algorithm for estimating commonly used is divided into two classes: partial estimation is estimated with the overall situation.In this patent, adopt overall algorithms to estimate the intrinsic dimension of Radio Map is estimated, and as the input variable of LDE algorithm.Adopt geodesic distance minimal spanning tree algorithm (Geodesic Minimum Spanning Tree, GMST) that the intrinsic dimension of Radio Map is estimated in this patent.The below analyzes the theory of GMST algorithm.

Geodesic curve minimum spanning tree (GMST) estimates that the length function that is based on the geodesic curve minimum spanning tree depends on intrinsic dimension d's strongly.GMST refers to be defined in the minimum spanning tree of the neighbour's curve on the data set X.The length function of GMST is the Euclidean distance sum that all edges are corresponding in the geodesic curve minimum spanning tree.

Similar to ISOMAP, GMST estimates to construct neighbour's curve G at data set X, wherein, and each data point x in X _iAll with its k neighbour

Be connected.Geodesic curve minimum spanning tree T is defined as the minimum curve on the X, and it has length:

Wherein, Υ is all subtree collections of curve G, and e is the edge of tree T, g _eEuclidean distance corresponding to edge e.In GMST estimates, some subsets

M forms by all size, and the length L of the GMST of subset A (A) also needs to calculate.

In theory,

Linear, thus can by: the function of this form of y=ax+b is estimated, can estimate variable a and b by least square method.

Can prove: by the estimated value of a and

Can access the estimation of intrinsic dimension.

The expression formula that is provided intrinsic dimension d by the GMST algorithm is shown in the formula (2).Intrinsic dimension d is another important input parameter of LDE algorithm.

$d=\frac{1}{1-a}---\left(2\right)$

The realization of utilization LDE algorithm is carried out dimensionality reduction to Radio Map and is obtained feature weight matrix process and realize by following step:

The LDE algorithm is based on the maximized a kind of manifold learning arithmetic of divergence in class scatter and the class.Before the LDE algorithm is carried out theory analysis, the given input data of LDE algorithm are done following explanation: input high dimensional data point

Data point x _iClass be labeled as y _i∈ 1,2 ..., P}, wherein P represents high dimensional data is divided into P submanifold, and the high dimensional data that is about to input is divided into the P class.

The high dimensional data of input is expressed as the form of matrix: X=[x ₁, x ₂..., x _m] ∈ R ^{N * m}From the form of matrix notation, the row in the matrix represent a high dimensional data point.Below in conjunction with mistake! Do not find Reference source.Shown LDE algorithm flow comes its theory of algorithm is derived.

The structure adjacent map: class label information and neighbor relationships thereof according to high dimensional data point are constructed directionless figure G and G '.Wherein neighbor relationships is the criterion that adopts the KNN algorithm to provide, and namely selects the nearest K of data point point as its neighbours, and G represents to work as x _iWith x _jClass label information y _i=y _jThe time and x _i, x _jK nearest neighbor concerns each other; G ' shows and works as x _iWith x _jClass label information y _i≠ y _jThe time and x _i, x _jK nearest neighbor concerns each other.

Calculate weight matrix: the adjacent map according to (1) structure adopts the class Gaussian function to carry out the calculating of weight matrix.Its expression formula is shown in (3).W in the formula (3) _IjExpression Neighbor Points x _iWith x _jBetween weights, || x _i-x _j|| ²Be Neighbor Points x _iWith x _jBetween the norm distance, adopt matrix-style to calculate the norm distance, t is the weights normalized parameter.

$\left\{\begin{array}{c}{w}_{\mathrm{ij}}=\left\{\begin{array}{c}\exp (-{\left|\right|x_i-x_j\left|\right|}^2/t);\\ 0;\end{array}\mathrm{if}{x}_i,x_j\∈G\right.\\ w_{\mathrm{ij}}^{\′}=\left\{\begin{array}{c}\exp (-{\left|\right|x_i-x_j\left|\right|}^2/t);\\ 0;\end{array}\mathrm{if}{x}_i,x_j\∈G^{\′}\right.\end{array}\right.---\left(3\right)$

Calculate and embed the result: according to the target of LDE algorithm---divergence in the maximization class scatter ground simultaneous minimization class.Divergence adopts expression like numbers strong point and inhomogeneous norm apart from expression.Target by the LDE algorithm can draw its corresponding optimization aim function, suc as formula shown in.

$\left\{\begin{array}{c}\mathrm{MaximizeJ}\left(V\right)=\underset{i,j}{\mathrm{\Σ}}{\left|\right|V^T{x}_i-V^T{x}_j\left|\right|}^2{w}_{\mathrm{ij}}^{\′}\\ \mathrm{subjectto}\underset{i,j}{\mathrm{\Σ}}{\left|\right|V^T{x}_i-V^T{x}_j\left|\right|}^2{w}_{\mathrm{ij}}=1\end{array}\right.---\left(4\right)$

Do following the analysis according to the optimization aim function that formula (4) provides:

Calculating formula according to matrix norm:

Calculating formula is expressed as the computational methods of the matrix norm of matrix A, and the method that calculating formula provides is consistent with the calculating formula of matrix trace, that is: || and A|| ²=tr (AA ^T).Formula (4) can be expressed as the account form of matrix trace thus:

$J\left(V\right)=\underset{i,j}{\mathrm{\Σ}}\left\{\mathrm{tr}\right[(V^T{x}_i-V^T{x}_j){(V^T{x}_i-V^T{x}_j)}^T{w}_{\mathrm{ij}}^{\′}\left]\right\}---\left(5\right)$

Formula (5) can be reduced to:

$J\left(V\right)=\underset{i,j}{\mathrm{\Σ}}\left\{\mathrm{tr}\right[V^T(x_i-x_j){(x_i-x_j)}^{T}V\left]w_{\mathrm{ij}}^{\′}\right\}---\left(6\right)$

Scalar nature and weights element by the calculating of trace of a matrix are real number, formula (6) can be reduced to:

$J\left(V\right)=\mathrm{tr}\left\{V^T\underset{i,j}{\mathrm{\Σ}}\right[(x_i-x_j)w_{\mathrm{ij}}^{\′}({x_i}^T-{x_j}^T)\left]V\right\}---\left(7\right)$

According to simple mathematical relationship, formula (7) can be reduced to:

J(V)＝2tr{V ^T[X(D′-W′)X ^T]V} (8)

In the formula (8): X is the input data, and λ, v are eigen vector, and W, W ' are respectively the weight matrix of G and G ' correspondence, and D and D ' are diagonal matrix, and its diagonal element can be represented by formula (9).

$\left\{\begin{array}{c}{d}_{\mathrm{ii}}=\underset{j}{\mathrm{\Σ}}{w}_{\mathrm{ij}}\\ d_{\mathrm{ii}}^{\′}=\underset{j}{\mathrm{\Σ}}{w}_{\mathrm{ij}}^{\′}\end{array}\right.---\left(9\right)$

According to the derivation mode of formula (8), in like manner the constraints in (4) can be write as suc as formula (8) similar form, thus, (4) can be expressed as form:

$\left\{\begin{array}{c}\mathrm{MaximizeJ}\left(V\right)=2\mathrm{tr}\left\{V^T\right[X(D^{\′}-W^{\′})X^T\left]V\right\}\\ \mathrm{subjectto}2\mathrm{tr}\left[X(D-W)X^T\right]=1\end{array}\right.---\left(10\right)$

Formula (10) is used Lagrange (Lagrange) Multiplier Method, can draw shown in the formula (11):

X(D′-W′)X ^Tv＝λX(D-W)X ^Tv (11)

Formula (11) is carried out generalized eigenvalue decomposition, draw characteristic value and the characteristic vector of its Eigenvalues Decomposition, be expressed as: λ=[λ ₁, λ ₂..., λ _n] ^T, its characteristic of correspondence vector is: v=[v ₁, v ₂..., v _n] ^TGet front d maximum characteristic value characteristic of correspondence vector and consist of transformation matrix V=[v ₁, v ₂..., v _d].Output data conversion method by the LDE algorithm can draw, and data are behind the dimensionality reduction:

z _i＝V ^Tx _i (12)

In the formula (12), z _iExpression input high dimensional data point x _iLow-dimensional output data after the conversion.

Above-mentioned analysis is theory analysis and the explanation that provides according to the LDE algorithm flow.

The theory that is provided the LDE algorithm by formula (5)～(11) is derived.Can draw Radio map and eigentransformation matrix behind the dimensionality reduction by the LDE algorithm, be designated as respectively Radio Map ^*And V '.

Online positioning stage is realized by following step RSS and KNN coupling location:

In conjunction with the flow chart shown in the on-line stage of Fig. 2 embodiment four is elaborated.On-line stage, the RSS=[AP that the test point place receives ₁, AP ₂..., AP _n], with eigentransformation matrix V ' multiply each other, thereby draw RSS ' behind the dimensionality reduction=[AP ₁, AP ₂... AP _d], wherein d represents the intrinsic dimension.Adopt again the KNN algorithm to realize RSS ' and Radio Map ^*Coupling.Adopt with the mean value of the coordinate of nearest K the reference point of RSS ' as test point (x ', y '), its expression formula is:

$(x^{\′},y^{\′})=\frac{1}{K}\underset{i=1}{\overset{K}{\mathrm{\Σ}}}(x_i,y_i)---\left(13\right)$

AP concrete indoor layout and experimentation in the following example shown in: Fig. 1 is shown the plane graph signal of the research park 2A of Harbin Institute of Technology 12 floor, just is based under this experimental situation based on the indoor locating system of WiFi and sets up.In experimental situation, altogether arrange 27 AP, AP is 2 meters from the room ground level.In off-line phase, at the V450 of association notebook NetStumbler software is installed, 100 RSS values of continuous sampling record AP on four of all reference points different orientation, and the relevant information of AP.The physical coordinates of all sampled points and corresponding physical coordinates and RSS value are stored as the data that position fixing process calls, set up Radio Map.Have 900 reference points in experimental situation, its sampling density is 0.5 meter * 0.5 meter.Radio Map is as the input data of input parameter and the intrinsic dimension algorithm for estimating of LDE algorithm.

Claims (7)

1. based on the WiFi indoor positioning method of LDE algorithm, it is characterized in that: it is realized by the following steps: Step 1. Arranging N access points APs for the indoor environment, ensuring that any point in the environment is covered by signals sent by one or more access points APs, and the N access points APs form a WiFi network; Set N _RP reference points evenly in the indoor environment; N and N _RP are both positive integers; Step 2: Select a reference point to establish a two-dimensional rectangular coordinate system as the coordinate origin, obtain the coordinate positions of N _RP reference points in the two-dimensional rectangular coordinate system, and use a signal receiver to collect on each reference point in the offline stage The signal strength RSS value from each access point AP is used as the location characteristic information of the access point AP; and the indoor signal coverage map Radio Map is constructed according to the location characteristic information of the N access point APs; Step 3, using the eigendimension estimation algorithm to perform eigendimensional analysis on the indoor signal coverage map Radio Map obtained in step 2, and obtain the eigendimensional analysis results; Step 4. According to the eigendimensional analysis results obtained in step 3, use the LED algorithm to reduce the dimensionality of all reference points in the indoor signal coverage map RadioMap to the eigendimensionality, obtain the feature transformation matrix, and generate the signal coverage map after dimensionality reduction Radio Map ^* ; Step 5, in the online phase, measure the signal strength RSS value of the desired positioning point in the indoor environment, and multiply the signal strength RSS value with the characteristic transformation matrix obtained in step 4 to obtain the signal strength transformation value RSS ^* ; Step 6. Use the KNN algorithm to match the signal strength transformation value RSS ^* obtained in step 5 with the reduced-dimensional signal coverage map Radio Map ^* generated in step 4 to obtain the position coordinates of the desired positioning point and complete the indoor location of the desired positioning point. position. 2. The WiFi indoor positioning method based on the LDE algorithm according to claim 1, characterized in that in the step 3, the eigendimension estimation algorithm is adopted as the eigenvalue estimation method. 3. The WiFi indoor positioning method based on the LDE algorithm according to claim 1, characterized in that in step 3, the intrinsic dimension estimation algorithm is adopted as the packet number estimation method. 4. The WiFi indoor positioning method based on the LDE algorithm according to claim 1, characterized in that in the step 3, the eigendimension estimation algorithm is adopted as the geodesic minimum spanning tree algorithm. 5. the WiFi indoor positioning method based on LDE algorithm according to claim 4, it is characterized in that adopting geodesic minimum spanning tree algorithm to carry out eigendimensional analysis to the indoor signal coverage figure Radio Map that step 2 obtains is by formula: ${\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; intrinsic\; dim</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; a</span>}}$ Realized; where: d _intrinsicdim is the result of intrinsic dimension analysis; a represents the slope of the linear fitting expression of the minimum spanning tree. 6. the WiFi indoor positioning method based on LDE algorithm according to claim 1, it is characterized in that in the step 4, the relationship between the feature transformation matrix V ' and the signal coverage map Radio Map ^* after generating dimensionality reduction is: Radio Map ^* = V'·Radio Map. 7. The WiFi indoor positioning method based on the LDE algorithm according to claim 1, characterized in that the signal strength conversion value RSS ^* obtained in step 5 and the signal coverage map RadioMap ^* after dimension reduction generated in step 4 are carried out by using the KNN algorithm The method of location matching is through the formula: $<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; K</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; K</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$ achieved; In the formula: (x', y') is the coordinate of the point to be located, ( _xi , _y ) is the coordinate of the i-th neighbor point, i is a positive integer; K is the total number of neighbor points in the KNN algorithm.

Images