CN104519571A - Indoor positioning method based on RSS (Received Signal Strength) - Google Patents

Abstract

The invention discloses an indoor positioning method based on RSS (Received Signal Strength). The data acquisition is convenient, extra receiving equipment is not needed, and the positioning is accurate. The indoor positioning method based on RSS comprises the following steps: (1) at an offline stage, collecting wireless signal receiving strength information on some positions of a space to construct a fingerprint database; (2) at an online stage, collecting signal strength on a walking path; (3) adding time and space constraint conditions by using a sparse representation algorithm, and establishing a positioning model; (4) calculating position coordinates corresponding to signal values on the path, and optimizing the result.

Description

A kind of indoor orientation method based on RSS

Technical field

The invention belongs to WLAN (Wireless Local Area Networks, WLAN) technical field of indoor positioning, relate to a kind of indoor orientation method based on RSS (Received signal strength, received signal strength) particularly.

Background technology

WLAN (WLAN) is a kind of brand-new information acquisition platform, can realize complicated location on a large scale, monitoring and tracking task, and network node self poisoning is basis and the prerequisite of great majority application in application widely.Current popular Wi-Fi (Wireless Fidelity, adopting wireless fidelity technology) location is a kind of location solution of the IEEE802.11 of WLAN series standard.The mode that this system adopts experience test and signal propagation model to combine, be easy to install, need little base station, can adopt identical bottom wireless network architecture, system overall accuracy is high.Mainly be divided into following three classes:

● approximation method

Approximation method utilizes AP in the feature (router of different model has different coverages) of the limited coverage area of indoor, is judged the position of mobile subscriber by the situation of terminal equipment received signal strength and the position of corresponding A P.When user is near a certain known location, carry out positioning object by this position.Namely, wireless terminal is used for the position of the accessing points (AP) of data communication, approx as the position estimated.The method can be used for test item contact, the aspects such as monitors for cellular network access point.It does not need complicated calculating, but positioning precision is confined to the coverage of AP, can only the location determination of feasible region property, and needs the prior information of AP particular location.

● geometric measurement method

First this method requires the propagation model (empirical model or Mathematical Modeling) according to radio signal, signal strength values is mapped as the distance that signal is propagated.On two dimensional surface, according to the distance between terminal equipment and other at least three AP, carry out location estimation by the geometry principle of trilateration.As shown in Figure 1, three black round dots are the reference point of known coordinate, and x is the point needing location, then utilize the distance of x and three reference point, can calculate the coordinate of x.

RADAR (the radio detection and ranging of Microsoft, radio detection and ranging) be an indoor locating system based on RSSI (the signal strength signal intensity instruction that Received Signal Strength Indication receives) technology, also be the indoor locating system based on WLAN occurred the earliest, make full use of existing WLAN facility, determine the position of user node in floor by the received signal strength indicator in 802.n standard.Usual use two kinds of method computing node positions, wherein a kind of is the theoretical model utilizing signal to propagate.This precision of method is not high, but can cost saving, need not shift to an earlier date building database, after base station movement, recalculate parameter frequently.But in actual environment, the conditions such as temperature, barrier, communication mode are all often changes, make this technology still have difficulties in actual applications.These class methods are simple, and computational efficiency is high, but whether correctly the accuracy of location depends on propagation model, whether is applicable to the building structure of locating area complexity.Due to the complexity of indoor electric wave traveling, signal strength signal intensity is subject to the impact such as multipath transmisstion, reflection, makes to be difficult to portray by fixing Mathematical Modeling in actual indoor environment.

● scene analysis method

Scene analysis method, be also called fingerprint technique, not directly the measurement of signal strength values is mapped as signal propagation distance, but utilize the scene characteristic observed in a certain place to infer the position of observer, can be regarded as and first the inherent law between signal strength signal intensity and position is learnt, and then mate with the sample point learnt with new measured value.

The method is generally divided into off-line measurement and two stages of tuning on-line to complete location.Off-line measurement selectes some sample points according to certain spacing distance in the region needing location, form the grid of a sample point, these sample point position are measured, records the signal strength measurement vector from each AP, these information structures signal strength signal intensity fingerprint base.This fingerprint base describes the relation of signal strength signal intensity and locus in this stationary positioned environment.Tuning on-line is compared with the information in fingerprint base by the signal strength measurement measured in real time vector, gets the position of the immediate sample point of signal strength signal intensity as the position estimated.

The method of the another kind of computing node position in RADAR system, utilizes the empirical model that signal is propagated exactly.Before actual location, in floor, choose some test points, record these and put the signal strength signal intensity that last base station receives, set up the offline database of position and signal strength signal intensity relation on each aspect.During actual location, system adopts certainty matching algorithm K nearest-neighbors (KNN) algorithm one by one, signal strength signal intensity according to recording in the signal strength signal intensity recorded and database compares, and the coordinate of that point that signal strength signal intensity mean square deviation is minimum is the coordinate of node.This method has higher precision, but will set up position and signal strength signal intensity relational database in advance, will re-establish database when base station movement.

This system can embed on the hand-held terminal device of any Wifi of having adapter, and independently positions and follow the trail of, and without the need to extra hardware supports, and without the need to line-of-sight transmission, thus comparatively Cricket system is wide for orientation range.But due to the complexity of indoor environment, as multipath, shadow fading, interference etc., indoor electric wave traveling has stronger time-varying characteristics, the performance of locating is made to be subject to certain impact.

Because radio waves propagation model that fingerprint technique is more traditional can describe the relation of RSS and locus more accurately, and without the need to the prior information of AP particular location, be thus widely used in the indoor locating system based on RSS.

Summary of the invention

Technology of the present invention is dealt with problems and is: overcome the deficiencies in the prior art, provides a kind of indoor orientation method based on RSS, and its data acquisition facilitates, without the need to additionally increasing receiving equipment, accurate positioning.

Technical solution of the present invention is: this indoor orientation method based on RSS, comprises the following steps:

(1) off-line phase, the radio signal reception strength information on some positions of collection space, builds fingerprint base;

(2) the on-line testing stage, the signal strength signal intensity on walking path is collected;

(3) by using rarefaction representation algorithm, joining day and space constraints, set up location model;

(4) position coordinates that on calculating path, signal value is corresponding, is optimized result.

Rarefaction representation algorithm applies to set up location model by the present invention, and joining day and space constraints, so data acquisition is convenient, without the need to additionally increasing receiving equipment, location is more accurate.

Accompanying drawing explanation

Fig. 1 is the schematic diagram carrying out location estimation according to geometric measurement method.

Fig. 2 is real experiment scene vertical view.

Fig. 3 a, b, c, d are according to context of methods respectively, k nearest neighbor method, and rarefaction representation algorithm and kernel method carry out the schematic diagram of straight line path location in real scene.

Embodiment

This indoor orientation method based on RSS (Received signal strength, received signal strength), comprises the following steps:

(1) off-line phase, the radio signal reception strength information on some positions of collection space, builds fingerprint base;

(2) the on-line testing stage, the signal strength signal intensity on walking path is collected;

(3) by using rarefaction representation algorithm, joining day and space constraints, set up location model;

(4) position coordinates that on calculating path, signal value is corresponding, is optimized result.

Off-line phase: also claim the training stage, obtains the process of finger print data and structure fingerprint base;

Finger print data: the received signal strength data of space known location, refers to the signal strength data of the multiple wireless wifi node (AP) known location utilizing the mobile devices such as smart mobile phone obtain in this article;

Fingerprint base: the set of given space finger print data.Usually be grid by spatial division, record its finger print data to the position of each grid node, the finger print data of all nodes forms fingerprint base;

The on-line testing stage: in tested object moving process, utilize the signal strength signal intensity on the mobile device record move paths such as smart mobile phone, and adopt sparse representation model to realize the location estimation of mobile object;

Rarefaction representation algorithm: be a kind of method for expressing to signal, Setting signal, by a dictionary obtained in advance (i.e. the fingerprint base of this paper), is expressed as the linear combination of data in dictionary by the method.

Time constraint condition: object is in moving process, and the signal receiving strength that path is recorded is continually varying, therefore also has the continuity of time for the rarefaction representation of continuous signal;

Space constraints: wireless signal strength spatially has the characteristic of continuous distribution, namely for the signal receiving strength that position, one, space is measured, have continuity and similitude with the signal strength signal intensity that its peripheral location is measured, therefore the rarefaction representation of the signal strength signal intensity of certain position is also only relevant with the rarefaction representation of the position signalling intensity that this locational space closes on.

Rarefaction representation algorithm applies to set up location model by the present invention, and joining day and space constraints, so data acquisition is convenient, without the need to additionally increasing receiving equipment, location is more accurate.

Preferably, location model is obtained by formula (5)-(7) in described step (3):

$S={\left(\begin{array}{cccccc}1& 1& 0& ...& 0& 0\\ 1& 1& 1& ...& 0& 0\\ 0& 1& 1& ...& 0& 0\\ ...& ...& ...& ...& ...& ...\\ 0& 0& 0& ...& 1& 1\\ 0& 0& 0& ...& 1& 1\end{array}\right)}_{M\×N}---\left(5\right)$

$T={\left(\begin{array}{ccccc}1& 0& 0& ...& 0\\ -1& 1& 0& ...& 0\\ 0& -1& ...& ...& 0\\ 0& 0& ...& -1& 1\\ 0& 0& ...& 0& -1\end{array}\right)}_{n\×(n-1)}---\left(6\right)$

Wherein λ 1,2 (should be two threshold parameters) are the threshold value of setting, Y=[y ₁, y ₂... y _n] the continuous received signal strength that gathers in moving process for mobile object, y _irepresent the signal receiving strength vector of i-th time period collection, ψ is the fingerprint base in step (1), in each list show every column signal in Y rarefaction representation vector.Solve and obtain according to the signal location information in fingerprint base ψ, just can obtain the positional information of every column signal in Y, thus realize location, position.

Preferably, obtain by formula (8) result optimized in described step (4):

$\begin{array}{c}R=\left\{n\right|\widehat{\mathrm{\θ}}\left(n\right)>r\}\\ (\hat{x},\hat{y})=\frac{1}{{\mathrm{\Σ}}_{n\∈R}\widehat{\mathrm{\θ}}\left(n\right)}\underset{n\∈R}{\mathrm{\Σ}}(\widehat{\mathrm{\θ}}\left(n\right)\·(x_n,y_n))\end{array}---\left(8\right)$

Wherein r is threshold value, and R is θ _iin be greater than the location sets of threshold value, (x _n, y _n) represent at the coordinate figure of n point, for the weights in the n-th position.

Below provide a specific embodiment:

The present invention mainly comprises rarefaction representation and the restructing algorithm of signal.

Introduce the Mathematical Modeling of rarefaction representation below, one real-valued limit for length's one-dimensional discrete time signal x, can see a R as ⁿthe column vector of N × 1 dimension in space, element is x [n], n=1,2 ..., N.If image or high dimensional data vector, then change into a long one-dimensional vector.R ⁿany signal in space can use the orthogonal base vectors of N × 1 dimension linear combination represent.Vector the orthogonal basis dictionary matrix of N × N is formed as column vector arbitrary signal x can be expressed as

$\begin{array}{ccc}x=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\θ}}_i{\mathrm{\ψ}}_i& \mathrm{or}& x=\mathrm{\Ψ\θ}\end{array}$

Wherein θ is weight coefficient θ _i=<x, the column vector of N × 1 formed, () ^trepresent matrix transpose operation.Obviously, x and θ is the equivalent representation of same signal, and x is the expression of signal in real domain, and θ is then the expression of signal in Ψ territory.

Sparse signal representation basic thought is exactly the main information representing signal with the least possible non-zero coefficient, thus simplifies the solution procedure of signal processing problems.Existing sparse signal representation method can be divided into orthogonal basis rarefaction representation and redundant dictionary rarefaction representation two class.

Natural sign in usual time domain all right and wrong is sparse, but may be sparse at some transform domain.Orthogonal basis rarefaction representation make use of this feature of signal just, is projected to by signal on orthogonal transform base, obtains sparse or approximate sparse conversion vector.General Transformations base can be chosen flexibly according to the feature of signal itself, the total variation norm of the Fourier coefficient of smooth signal and wavelet coefficient, the functions of bounded variation, the Gabor coefficient of oscillator signal, the Curvelet coefficient etc. with the picture signal of discontinuous edge all have enough openness, can by the theoretical restoring signal of CS.If apply orthogonal basis sparse representation method herein, so how finding the orthogonal basis of applicable wifi signal, even how to construct the orthogonal basis of applicable wifi signal, in the hope of the most rarefaction representation of wifi signal, will be the key issue needing research.

When signal can not use orthogonal basis rarefaction representation, redundant dictionary rarefaction representation can be adopted.Recent years, another focus studied rarefaction representation is the Its Sparse Decomposition of signal under redundant dictionary. the super complete redundancy functions storehouse substituting group function of this theory, be referred to as redundant dictionary, element in dictionary is called as atom. and the selection of dictionary should meet as well as possible by the structure of approximation signal, its formation can without any restriction. and from redundant dictionary, find the K item atom with optimum linear combination to represent a signal, the sparse bayesian learning or the nonlinearity that are called signal approach.If adopt redundant dictionary sparse representation method herein, the emphasis that following two problems will be research: how (1) constructs the redundant dictionary of an applicable wifi signal; (2) Its Sparse Decomposition algorithm fast and effectively how is designed.

For research method herein, adopt the method for redundant dictionary to build ψ, concrete process, supposes off-line training, for all N number of AP, gathers the signal strength signal intensity of N number of fingerprint point, can obtain the matrix of a N*N, represent i-th fingerprint point, each be all N*1 dimension, represent the measured value of N number of AP.Use the sparse vector θ that compressed sensing draws, each row just represents those row of corresponding fingerprint base, and the value of element just represents the influence degree of these row for signal strength signal intensity.The respective column of fingerprint base just represents a fingerprint point, a coordinate points, just can obtain the corresponding coordinate position of test signal like this, finally obtaining positioning result by obtaining θ.

● measuring-signal y

Measuring-signal y _jrepresent the vector of all AP all Signal reception values within a jth time period.P _i,jrepresent the measured value of i-th AP within a jth time period.N is the number of AP.Obtain following formula:

$y_j{=[p_{1,j},p_{2,j},...p_{N,j}]}_{N\&times;1}^T---\left(2\right)$

According to formula (3), can solve and obtain θ, θ is representative represents signal y sparse vector with the least possible non-zero coefficient.The method of rarefaction representation is used to solve positional information in this article, y is such as formula (2), ψ is the signal strength signal intensity of all fingerprints point, according to formula (3), solve θ, here i-th element in θ, correspond to the i-th column signal intensity level in ψ, is equivalent to corresponding i-th coordinate points.Therefore by solving θ, position coordinates point can just be obtained.

$\begin{array}{ccc}\widehat{\mathrm{\&theta;}}=\arg \min {\left|\right|\mathrm{\&theta;}\left|\right|}_1& s.t.& y=\mathrm{\&psi;\&theta;}\end{array}----\left(3\right)$

At test phase, gather one section of continuous signal Y, Y is divided into n section, obtains Y=[y ₁, y ₂... y _n], y _ithe vector of representative all AP Signal reception values within i-th period of running time.Present fingerprint base ψ, by sparse representation model, as (4), can obtain in comprise a series of θ _i, namely the sparse vector solution of every column signal in Y is shown in each list in it, θ _isignal location information in corresponding ψ, so just obtains the positional information of every column signal in Y, thus realizes the location of all positions.

Existing location algorithm is just compared to single signal and is obtained last position, for spatially too much not paying close attention to the relation of temporal signaling point.Signal message collected by training stage, spatially, the Influence on test result of point around anchor point to location should be larger, signal message corresponding on time should be associated in space, herein by rarefaction representation, obtain the positional information of every bit, on this basis constraint is added for point and the positional information of point, finally realize location.

For each fingerprint point, the signal strength signal intensity of the fingerprint point of its surrounding should be close with it, and position should be more close, and the impact of surrounding point should be larger, and in this article, in sparse vector, the value of correspondence position should be larger, for the fingerprint base with N number of fingerprint point, builds space constraints, S as shown in the formula

$S={\left(\begin{array}{cccccc}1& 1& 0& ...& 0& 0\\ 1& 1& 1& ...& 0& 0\\ 0& 1& 1& ...& 0& 0\\ ...& ...& ...& ...& ...& ...\\ 0& 0& 0& ...& 1& 1\\ 0& 0& 0& ...& 1& 1\end{array}\right)}_{M\&times;N}---\left(5\right)$

answer the information constrained threshold value of meeting spatial.

For the signal value that same AP is corresponding, the signal message corresponding with more lower position of the upper position in walking differs should be smaller, so the θ obtained _iwith θ _jbetween distance should be more or less the same.For the walking path with n location point, build such constraints, T is as (6) formula

$T={\left(\begin{array}{ccccc}1& 0& 0& ...& 0\\ -1& 1& 0& ...& 0\\ 0& -1& ...& ...& 0\\ 0& 0& ...& -1& 1\\ 0& 0& ...& 0& -1\end{array}\right)}_{n\&times;(n-1)}---\left(6\right)$

temporal information constraint threshold value should be met.

To sum up, model is obtained as follows:

λ 1,2 is the threshold value of setting, finally solves according to above formula obtain final positioning result.

Because θ is not only 1 sparse vector, so arrange a threshold value r, last position calculation is made in some positions that value is larger than r, as follows:

$\begin{array}{c}R=\left\{n\right|\widehat{\mathrm{\&theta;}}\left(n\right)>r\}\\ (\hat{x},\hat{y})=\frac{1}{{\mathrm{\&Sigma;}}_{n\&Element;R}\widehat{\mathrm{\&theta;}}\left(n\right)}\underset{n\&Element;R}{\mathrm{\&Sigma;}}(\widehat{\mathrm{\&theta;}}\left(n\right)\&CenterDot;(x_n,y_n))\end{array}---\left(8\right)$

The method mainly under actual application environment, the fixation and recognition problem of people's walking position.Utilize the method for rarefaction representation to carry out the calculating of location point weights.People is in the process of walking, and each location point and a upper location point distance are close, so joining day constraints, limit the distance between each location point; Each location point only with it around several fingerprint points have relation, so add the weights size that space constraint limits each fingerprint point.

Illustrate an actual environment embodiment below.

One, actual environment embodiment

1. the foundation of actual environment and the sampling of signal strength data

In indoor true environment, experimental site is located at building three, Beijing University of Technology's information north floor, long 53 meters, wide 15 meters, as shown in Figure 2.In this experiment, we sample the RSS value from 90 AP altogether in this region.Off-line training step one people carries mobile terminal and walks in Experimental Area, records RSS value and coordinate simultaneously.210 sampled points are gathered altogether in experiment.In order to avoid systematic error, obtain accurate measured value, each sampled point we all carried out 10 times sampling.The average of 10 samples is registered as the final measured value of this sampled point.

2. design walking path

In the on-line testing stage, a people carries mobile terminal and walks in Experimental Area, devises straight line path, obtains the signal strength signal intensity of all location points on path.

3. add time and space constraint matrix, positioned by the rarefaction representation restricted model of structure

Mainly sparse vector be with the addition of to the constraint of the time and space in the present invention, time-constrain matrix major embodiment be constraint between current location and next position, space constraint mainly retrains the position of current point, and the point around it should be maximum on its impact.The signal strength signal intensity of test vector uses the model adding the rarefaction representation algorithm retrained to calculate sparse vector with the fingerprint base constructed by positioning stage.

4. obtain position coordinates by sparse vector

Do product calculation by sparse vector and fingerprint base coordinate and obtain tuning on-line coordinate.

Experimental result:

Error (rice)	Context of methods	K nearest neighbor method	Rarefaction representation algorithm	Kernel method
					Straight line path	1.1284m	1.5214m	1.5339m	1.2859m

Error is the difference between true path from the positioning result using different localization methods to obtain.K nearest neighbor method, rarefaction representation algorithm and kernel method are the location algorithms based on signal strength signal intensity relatively commonly used.Experiment scene is the space of a 53m*15m, error Shi meter office.Unit in accompanying drawing is centimetre, and black circle is real walking path, and on curve, asterism is the position using algorithm to obtain.

The above; it is only preferred embodiment of the present invention; not any pro forma restriction is done to the present invention, every above embodiment is done according to technical spirit of the present invention any simple modification, equivalent variations and modification, all still belong to the protection range of technical solution of the present invention.

Claims (3)

1. based on an indoor orientation method of RSS, it is characterized in that, comprise the following steps:

(1) off-line phase, the radio signal reception strength information on some positions of collection space, builds fingerprint base;

(2) the on-line testing stage, the signal strength signal intensity on walking path is collected;

(3) by using rarefaction representation algorithm, joining day and space constraints, set up location model;

(4) position coordinates that on calculating path, signal value is corresponding, is optimized result.

2. according to claim 1 based on RSS (Received signal strength, received signal strength) indoor orientation method, it is characterized in that, in described step (3), obtain location model by formula (5)-(7):

$S={\left(\begin{array}{cccccc}1& 1& 0& ...& 0& 0\\ 1& 1& 1& ...& 0& 0\\ 0& 1& 1& ...& 0& 0\\ ...& ...& ...& ...& ...& ...\\ 0& 0& 0& ...& 1& 1\\ 0& 0& 0& ...& 1& 1\end{array}\right)}_{N\&times;N}---\left(5\right)$

$T={\left(\begin{array}{ccccc}1& 0& 0& ...& 0\\ -1& 1& 0& ...& 0\\ 0& -1& ...& ...& 0\\ 0& 0& ...& -1& 1\\ 0& 0& ...& 0& -1\end{array}\right)}_{n\&times;(n-1)}---\left(6\right)$

Wherein λ 1,2 is the threshold value set, Y=[y ₁, y ₂... y _n] the continuous received signal strength that gathers in moving process for mobile object, y _irepresent the signal receiving strength vector that i-th time point gathers, ψ is the fingerprint base in step (1), in each list show every column signal in Y rarefaction representation vector; Solve and obtain according to the signal location information in fingerprint base ψ, obtain the positional information of every column signal in Y, thus realize location, position.

3. according to claim 2 based on RSS (Received signal strength, received signal strength) indoor orientation method, it is characterized in that, in described step (4), obtained the result optimized by formula (8):

$R=\left\{n\right|\widehat{\mathrm{\&theta;}}\left(n\right)>r\}$

$(\hat{x},\hat{y})=\frac{1}{{\mathrm{\&Sigma;}}_{n\&Element;R}\widehat{\mathrm{\&theta;}}\left(n\right)}\underset{n\&Element;R}{\mathrm{\&Sigma;}}(\widehat{\mathrm{\&theta;}}\left(n\right)\&CenterDot;(x_n,y_n))---\left(8\right)$

Wherein r is threshold value, and R is θ _iin be greater than the location sets of threshold value, (x _n, y _n) represent at the coordinate figure of n point, for the weights in the n-th position.