CN103139907B - A kind of indoor wireless positioning method utilizing fingerprint technique - Google Patents

Abstract

The present invention discloses a kind of indoor wireless positioning method utilizing fingerprint technique greatly reducing requirement to sampling point density, enrich signal intensity profile information, keep high-precision positioning result while reducing workload, comprise off-line training step and tuning on-line stage, off-line training step comprises the following steps: (1) sets up indoor environment and collection signal intensity data; (2) the original fingerprint storehouse of low sampling rate is constructed; (3) by low-rank matrix loaded with dielectric, original fingerprint storehouse being reconstructed, does is low-rank matrix loaded with dielectric minrank (X)? s.t.A (X)=B; In the tuning on-line stage, mated by the signal strength signal intensity vector of the signal strength signal intensity being used for testing vector with each sampled point in the fingerprint base of reconstruct, current position coordinates utilizes each known sample point coordinate to carry out estimation to obtain tuning on-line coordinate.

Description

A kind of indoor wireless positioning method utilizing fingerprint technique

Technical field

The invention belongs to the technical field of wireless location, relate to a kind of indoor wireless positioning method utilizing fingerprint technique particularly.

Background technology

At present, location-based service and a series of application brought thus more and more receive the concern of people, and the demand of mobile subscriber's location information property and instantaneity on the spot increases day by day.The interested event of a lot of user, as environmental monitoring, logistics management, condition of a fire report, must combine with positional information and just have value.Localization method comparatively ripe at present has GPS technology, and be widely used in outdoor positioning, positioning precision can reach about 10 meters.And in indoor, due to the impact of the factors such as wall, positioning precision is difficult to the needs reaching people.Therefore at some specific areas, be the problem that people must consider as obtained high-precision positioning result in indoor environment.

Location based on wifi signal strength signal intensity has widely distributed, obtains the advantages such as convenient, has become one of focus of indoor wireless positioning method research.The method needs to set up multiple wireless router at interested locating area, as accessing points (AP).Each wireless router can send signal, and generally signal strength values (RSS) can be decayed along with the increase of distance.Different positions can receive the signal that different AP sends, and the signal strength values of reception is also different.Therefore, the signal strength values from each AP that can receive according to a certain position based on the location of wifi signal strength signal intensity positions.

Localization method based on wifi signal strength signal intensity is mainly divided into geometric measurement method and scene analysis method two kinds.First geometric measurement 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.But due to the complexity of indoor electric wave traveling, signal strength signal intensity is subject to the impact such as multipath transmisstion, reflection, make to be difficult to portray by fixing Mathematical Modeling in actual indoor environment.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 proposed in 2000 in the RADAR system of Microsoft, was generally divided into off-line measurement and two stages of tuning on-line to complete location.Off-line measurement selectes some sampled 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.During tuning on-line, 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 as the position estimated.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.

Setting up fingerprint base is the basis realizing positioning function.In order to reduce the impact that RSS instability is brought, traditional method setting up fingerprint base is the correlation utilizing the time, and under same sample point, repetitive measurement is averaged, as shown in Figure 1, each sample point collects the signal strength values of each AP, directly by these information stored in fingerprint base.Shih-HauFang proposes a kind of dynamical system, and the time series of RSS sample is merged into a kind of state, and using state replaces RSS directly to carry out location estimation.C.Feng adds directional information in fingerprint base, samples respectively, set up the fingerprint base on four direction to 0 °, 90 °, 180 °, 270 °, to reduce the factors such as the hand-held sample devices of people to the impact of signal.

Xing-chuanLiu proposes signal except time correlation, spatially also has correlation, the sampled data in certain radius is weighted on average, try to achieve a reference point.

BinghaoLi compares the quantity of sample point by experiment, shows that sample point interval reduces, and the precision of location estimation increases along with sample point increases.But sample point is more, linear growth can be brought to surveying work amount.Author points out sample point Existential Space correlation, and namely when measuring sub-fraction sample point, they provide not only the information under these positions, also provide the information of peripheral region, utilize spatial coherence can obtain more sample point information easily.Author have employed inverse distance weighted interpolation method method (LDW) and general Kriging regression method (UK) sets up fingerprint base, as shown in Figure 1, utilize adjacent sample point to collect the signal strength values of each AP, the signal strength values of the sample point between estimation, in the lump stored in fingerprint base.

Mostly the data of sampling are directly built up fingerprint base after statistics in existing indoor orientation method, but the density of sampled point and positioning precision are contacted directly, sampled point is more intensive, and positioning precision is higher.This causes wanting to obtain the high locating effect of precision, and off-line phase hand labor workload is large, inefficiency.In addition a part of localization method adopts interpolation method to build fingerprint base, but this method, the accuracy of interpolation point information cannot ensure.

Summary of the invention

Technology of the present invention is dealt with problems and is: overcome the deficiencies in the prior art, the indoor wireless positioning method utilizing fingerprint technique providing a kind of and greatly reduce requirement to sampling point density, enrich signal intensity profile information, keep high-precision positioning result while reducing workload.

Technical solution of the present invention is: this indoor wireless positioning method utilizing fingerprint technique, comprises off-line training step and tuning on-line stage,

Off-line training step comprises the following steps:

(1) indoor environment is set up and collection signal intensity data;

(2) the original fingerprint storehouse of low sampling rate is constructed;

(3) be reconstructed original fingerprint storehouse by low-rank matrix loaded with dielectric, low-rank matrix fills mould

Type is formula (1)

minrank(X)s.t.A(X)=B（1）

Wherein A (X) is template operator, X is a calculation matrix comprising whole sampled point, each element X (i, j) sampled point (i is represented, j) receive the signal strength signal intensity from a certain accessing points AP, A (X)=B, only have the effective signal strength values of existence of the relatively sparse sampled point be actually measured, A (X) is defined as the matrix Q of formula (2)

Then formula (3) is obtained

B(i,j)=Q(i,j)X(i,j)（3）；

In the tuning on-line stage, mated by the signal strength signal intensity vector of the signal strength signal intensity being used for testing vector with each sampled point in the fingerprint base of reconstruct, current position coordinates utilizes each known sample point coordinate to carry out estimation to obtain tuning on-line coordinate.

This method is at off-line training step low sampling rate down-sampling, by low-rank matrix loaded with dielectric to the fingerprint base comprising sampled point signal strength signal intensity intensive in a large number in refactoring localization region, original fingerprint storehouse, then carry out mating with the signal strength signal intensity vector of each sampled point in the fingerprint base of reconstruct in the tuning on-line stage and obtain tuning on-line coordinate, so just greatly reduce the requirement to sampling point density, enrich signal intensity profile information, while reduction workload, keep high-precision positioning result.

Accompanying drawing explanation

Fig. 1 shows the schematic diagram setting up fingerprint base according to the interpolation method of prior art;

Fig. 2 shows according to the flow chart utilizing the emulation experiment embodiment of the indoor wireless positioning method of fingerprint technique of the present invention;

Fig. 3 shows according to the flow chart utilizing the actual environment EXPERIMENTAL EXAMPLE of the indoor wireless positioning method of fingerprint technique of the present invention.

Embodiment

This indoor wireless positioning method utilizing fingerprint technique, comprises off-line training step and tuning on-line stage,

Off-line training step comprises the following steps:

(1) indoor environment is set up and collection signal intensity data;

(2) the original fingerprint storehouse of low sampling rate is constructed;

(3) be reconstructed original fingerprint storehouse by low-rank (Low-rank, LR) matrix fill-in model, low-rank matrix loaded with dielectric is formula (1)

minrank(X)s.t.A(X)＝B（1）

Wherein A (X) is template operator, X is a calculation matrix comprising whole sampled point, each element X (i, j) sampled point (i is represented, j) receive the signal strength signal intensity from a certain accessing points AP, A (X)=B, only have the effective signal strength values of existence of the relatively sparse sampled point be actually measured, A (X) is defined as the matrix Q of formula (2)

Then formula (3) is obtained

B(i,j)=Q(i,j)X(i,j)（3）；

In the tuning on-line stage, mated by the signal strength signal intensity vector of the signal strength signal intensity being used for testing vector with each sampled point in the fingerprint base of reconstruct, current position coordinates utilizes each known sample point coordinate to carry out estimation to obtain tuning on-line coordinate.

This method is at off-line training step low sampling rate down-sampling, by low-rank matrix loaded with dielectric to the fingerprint base comprising sampled point signal strength signal intensity intensive in a large number in refactoring localization region, original fingerprint storehouse, then carry out mating with the signal strength signal intensity vector of each sampled point in the fingerprint base of reconstruct in the tuning on-line stage and obtain tuning on-line coordinate, so just greatly reduce the requirement to sampling point density, enrich signal intensity profile information, while reduction workload, keep high-precision positioning result.

Preferably, step (3) also comprises the low-rank (SmoothingLow-Rank by being with smoothing, SLR) matrix fill-in model is reconstructed original fingerprint storehouse, and the low-rank matrix loaded with dielectric of band smoothing is formula (4) minrank (X)+λ S (X) s.t.A (X)=B

（4）

Wherein S (X) is a successional smoothing factor of expression X, and the value of S (X) is less, and the continuity representing X is better, and λ is the coefficient of balance obtained by experiment.

Preferably, define S (X) by the difference in matrix horizontal and vertical direction, see formula (5)

$S\left(X\right)={\left|\right|D_x\left(X\right)\left|\right|}_F^2+{\left|\right|D_y\left(X\right)\left|\right|}_F^2---\left(5\right)$

Wherein D _x(X) be a size be N ₁× (N ₂-1) matrix, represents the difference in each element level direction in matrix X, sees formula (6)

D _x(i,j)=X(i,j+1)-X(i,j)（6）

D _y(X) be a size be (N ₁-1) × N ₂matrix, represent the difference of each element vertical direction in matrix X, see formula (7)

D _y(i,j)=X(i+1,j)-X(i,j)(7)

Operator represent the Frobenius norm of matrix, obtain by formula (8) the low-rank matrix loaded with dielectric being with smoothing like this

$\min \mathrm{rank}\left(X\right)+\mathrm{\λ}({\left|\right|D_x\left(X\right)\left|\right|}_F^2+{\left|\right|D_y\left(X\right)\left|\right|}_F^2)$ s.t.A(X)=B（8）。

Preferably, in step (3), singular value decomposition (SingularValueDecomposition, SVD) method is adopted to the solving of low-rank matrix loaded with dielectric of low-rank matrix loaded with dielectric and band smoothing, solved by alternating iteration, generate the fingerprint base of reconstruct.

Preferably, SVD method is:

Can be divided into by formula (9) decomposition is three matrixes

X=U∑V ^T（9）

Wherein U is a size is N ₁× N ₁unitary matrice, V is a size is N ₂× N ₂unitary matrice, ∑ is a size is N ₁× N ₂diagonal matrix, comprise the singular value σ of descending _k, be formula (10) by matrix X factorization

X=U∑V ^T=LR ^T（10）

Wherein L=U ∑ ^1/2, R=V ∑ ^1/2, the low-rank matrix loaded with dielectric of this belt transect smoothing is updated to formula (11)

$\min \mathrm{rank}\left({\mathrm{LR}}^T\right)+\mathrm{\λ}({\left|\right|D_x\left({\mathrm{LR}}^T\right)\left|\right|}_F^2+{\left|\right|D_y\left({\mathrm{LR}}^T\right)\left|\right|}_F^2)$ s.t.A(LR ^T)=B（11）

If L is a size is N ₁the matrix of × K, R is a size is N ₂the matrix of × K, K is by the value of the order pre-estimation of matrix X here, and the low-rank matrix loaded with dielectric of this belt transect smoothing is updated to formula (12)

$\min {\left|\right|L\left|\right|}_F^2+{\left|\right|R\left|\right|}_F^2+\mathrm{\λ}({\left|\right|D_x\left({\mathrm{LR}}^T\right)\left|\right|}_F^2+{\left|\right|D_y\left({\mathrm{LR}}^T\right)\left|\right|}_F^2)$ s.t.A(LR ^T)=B（12）

Consider that the signal strength values that mobile terminal receives is usually accurate not, and the matrix in scene is not in full conformity with low-rank characteristic, relaxes A (LR ^tthe constraints of)=B, the low-rank matrix loaded with dielectric of the band smoothing of conversion belt restraining is unconfinement model, sees formula (13)

$\min {\left|\right|L\left|\right|}_F^2+{\left|\right|R\left|\right|}_F^2+\mathrm{\η}{\left|\right|A\left({\mathrm{LR}}^T\right)-B\left|\right|}_F^2+\mathrm{\λ}({\left|\right|D_x\left({\mathrm{LR}}^T\right)\left|\right|}_F^2+{\left|\right|D_y\left({\mathrm{LR}}^T\right)\left|\right|}_F^2)---\left(13\right)$

Wherein represent the reconstructed error of sampling subset B at balance weight η,

Derive L and R by above formula alternating iteration process: the first initial value of we random given L and R, then fixed L, optimizes R by least square method; Upgrade R afterwards, fixing R, allows L as optimized variable; Repeat above alternating iteration process, until objective function converges and reach default error threshold.

Illustrate an emulation experiment embodiment and an actual environment embodiment below.

One, emulation experiment embodiment

Fig. 2 is the flow chart of emulation experiment embodiment of the present invention, specifically comprises:

1. the first foundation of simulated environment and the generation of signal strength data

We are supposition random placement 50 AP in the long 100 meters wide rectangular areas of 50 meters.Then be the interval of a meter with horizontal and vertical step-length, design acquires the RSS value of 5000 sampled points altogether.In order to simulate the distribution spatially of RSS signal value, radio propagation path loss model be below used for simulate signal decay.

$P_r\left(d\right)=P_t\left(d\right)-\stackrel{\‾}{P}\left(d_0\right)-10n{\log }_{10}\left(\frac{d}{d_0}\right)-X_{\mathrm{\σ}}$

Wherein P _rthe position that d () represents in distance is d receives the signal strength signal intensity of AP, P _td signal strength signal intensity that () sends for AP, P (d ₀) represent apart from the position average signal strength loss value being d0, be generally 1 meter of.N is given value, is path loss index.X _σrepresent Gaussian noise distribution N (0, σ).When path loss index is known, RSS value can calculate.The furthest distance of disease spread of our putative signal is 30 meters, if the distance namely between AP and reference node is more than 30 meters, RSS value will be set as-100.In this experiment, path loss index n is set to 4.4 by us, and average signal strength loss (1m) is set to-35dB.Noise level limit is in [0,16] interval.According to above-mentioned experimental design, all reference nodes and AP can obtain.Finally, each AP is at 100 × 50 sampling point positions.These measured values form the above-mentioned original measurement matrix X mentioned.

2. construct the original fingerprint storehouse of low sampling rate

In fingerprint base reconstitution experiments, the sampled point of our usual Stochastic choice 20%, supposes that these sampled points are the sampled point of actual measurement.Namely form an incomplete calculation matrix B for each AP, wherein the element of 20% has valid value, represents the signal strength values receiving AP of this position, and other are 0.

3. carry out fingerprint base reconstruct by LR model and SLR model

Mainly original measurement matrix X is recovered out by LR model and SLR model two kinds of modes by approximate for incomplete calculation matrix B in the present invention.The method for solving that we are decomposed by SVD, alternating iteration derives L and R, until objective function converges and reach default error threshold.Merge the fingerprint distribution that each AP reconstructs, form new signal strength signal intensity fingerprint base.

4. utilize the signal strength signal intensity fingerprint base reconstructed to carry out tuning on-line

In the tuning on-line stage, mated with the signal strength signal intensity vector of each sampled point in the fingerprint base newly reconstructed by the signal strength signal intensity being used for testing vector, current position coordinates utilizes each known sample point coordinate to estimate, namely obtains tuning on-line coordinate

Two, actual environment embodiment

First 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.337 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. construct the original fingerprint storehouse of low sampling rate

Similar to emulation experiment, our usual Stochastic choice fractional-sample point, because do not carry out complete measument to the Experimental Area of 53m × 15m, we select a part as known sampled point at random from 337 reference nodes, namely an incomplete calculation matrix B is formed for each AP, wherein Partial Elements has valid value, represents the signal strength values receiving AP of this position, and other are 0.

3. carry out fingerprint base reconstruct by LR model and SLR model

Mainly original measurement matrix X is recovered out by LR model and SLR model two kinds of modes by approximate for incomplete calculation matrix B in the present invention.The method for solving that we are decomposed by SVD, alternating iteration derives L and R, until objective function converges and reach default error threshold.Merge the fingerprint distribution that each AP reconstructs, form new signal strength signal intensity fingerprint base.

4. utilize the signal strength signal intensity fingerprint base reconstructed to carry out tuning on-line

The tuning on-line stage, the signal strength signal intensity being used for measuring in real time vector is mated with the signal strength signal intensity vector of each sampled point in the fingerprint base newly reconstructed, current position coordinates utilizes each known sample point coordinate to estimate, namely obtains tuning on-line coordinate.

In order to verify the validity of above-mentioned structure fingerprint base method, we use emulated data and real data the present invention and prior art to be compared in fingerprint base structure result and positioning result two respectively.Fingerprint base builds the result evaluation that mainly people is visually subjective, and positioning result is mainly measured by objective position error, and unit is rice (m).Its computing formula is as follows:

$\mathrm{Error}=\left|\right|P-\hat{P}\left|\right|$

Wherein position error is actual position coordinate P and estimated position coordinate euclidean distance, position error is less, and locating effect is better.

Build result to show fingerprint base in said method intuitively, the wireless signal strength distribution of a certain AP represents with pseudocolour picture by we, and the color in image represents the signal strength values of this position.In emulation experiment, we simulate sample rate when being 20%, i.e. 1000 sampled points, utilize low-rank model reconstruction to go out the experiment of 5000 sampled points, reconstruction result and the signal distribution plots (IDW, RBF) that primary signal distributes and interpolation method obtains compare.We can find out that the signal distribution plots adopting LR method to carry out building is not ideal enough, and there is certain noise, the result adopting SLR method in this paper to carry out building is then more close with primary signal distribution map, reconstructs relatively better.In addition, based on being with the low-rank reconstructing method of smoothing than interpolation method, there iing better effect to the noise removed in environment, namely there is in noise circumstance better robustness.

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 (5)

1. utilize an indoor wireless positioning method for fingerprint technique, comprise off-line training step and tuning on-line stage, it is characterized in that,

Off-line training step comprises the following steps:

(1) indoor environment is set up and collection signal intensity data;

(2) the original fingerprint storehouse of low sampling rate is constructed;

(3) be reconstructed original fingerprint storehouse by low-rank matrix loaded with dielectric, low-rank matrix loaded with dielectric is formula (1)

Wherein A (X) is template operator, and X is a calculation matrix comprising whole sampled point, and each element X (i, j) represents sampled point (i, j) and receives signal strength signal intensity from a certain accessing points AP,

represent to only have on the relatively sparse sampled point be actually measured just there is effective signal strength values, sparse matrix represent the signal strength values of actual measurement, therefore A (X) operator is defined by the 0-1 sparse matrix Q in formula (2),

Namely the operator A from X to B is realized by the computing of the matrix element in formula (3):

B(i,j)＝Q(i,j)X(i,j)(3)

In the tuning on-line stage, mated by the signal strength signal intensity vector of the signal strength signal intensity being used for testing vector with each sampled point in the fingerprint base of reconstruct, current position coordinates utilizes each known sample point coordinate to carry out estimation to obtain tuning on-line coordinate.

2. the indoor wireless positioning method utilizing fingerprint technique according to claim 1, it is characterized in that, step (3) also comprises by being with the low-rank matrix loaded with dielectric of smoothing to be reconstructed original fingerprint storehouse, and the low-rank matrix loaded with dielectric of band smoothing is formula (4)

Wherein S (X) is a successional smoothing factor of expression X, and the value of S (X) is less, and the continuity representing X is better, and λ is the coefficient of balance obtained by experiment.

3. the indoor wireless positioning method utilizing fingerprint technique according to claim 2, is characterized in that, defines S (X) by the difference in matrix horizontal and vertical direction, sees formula (5)

Wherein be a size be N ₁× (N ₂-1) matrix, represents the difference in each element level direction in matrix X, sees formula (6)

D _x(i,j)＝X(i,j+1)-X(i,j)(6)

be a size be (N ₁-1) × N ₂matrix, represent the difference of each element vertical direction in matrix X, see formula (7)

D _y(i,j)＝X(i+1,j)-X(i,j)(7)

Operator represent the Frobenius norm of matrix, obtain by formula (8) the low-rank matrix loaded with dielectric being with smoothing like this

4. the indoor wireless positioning method utilizing fingerprint technique according to claim 3, it is characterized in that, in step (3), singular value decomposition SVD method is adopted to the solving of low-rank matrix loaded with dielectric of low-rank matrix loaded with dielectric and band smoothing, solved by alternating iteration, generate the fingerprint base of reconstruct.

5. the indoor wireless positioning method utilizing fingerprint technique according to claim 4, is characterized in that, SVD method is:

Can be divided into by formula (9) decomposition is three matrixes

X＝UΣV ^T(9)

Wherein U is a size is N ₁× N ₁unitary matrice, V is a size is N ₂× N ₂unitary matrice, Σ is a size is N ₁× N ₂diagonal matrix, comprise the singular value σ of descending _k, be formula (10) by matrix X factorization

X＝UΣV ^T＝LR ^T(10)

Wherein L=U Σ ^1/2, R=V Σ ^1/2, the low-rank matrix loaded with dielectric of this belt transect smoothing is updated to formula (11)

If L is a size is N ₁the matrix of × K, R is a size is N ₂the matrix of × K, K is by the value of the order pre-estimation of matrix X here, and the low-rank matrix loaded with dielectric of this belt transect smoothing is updated to formula (12)

Consider that the signal strength values that mobile terminal receives is usually accurate not, and the matrix in scene is not in full conformity with low-rank characteristic, relaxes constraints, the low-rank matrix loaded with dielectric of band smoothing of conversion belt restraining is unconfinement model, sees formula (13)

Wherein represent the reconstructed error of sampling subset B at balance weight η,

Derive L and R by above formula alternating iteration process: the first initial value of we random given L and R, then fixed L, optimizes R by least square method; Upgrade R afterwards, fixing R, allows L as optimized variable; Repeat above alternating iteration process, until objective function converges and reach default error threshold.