CN111757258A - Self-adaptive positioning fingerprint database construction method under complex indoor signal environment - Google Patents
Self-adaptive positioning fingerprint database construction method under complex indoor signal environment Download PDFInfo
- Publication number
- CN111757258A CN111757258A CN202010639947.2A CN202010639947A CN111757258A CN 111757258 A CN111757258 A CN 111757258A CN 202010639947 A CN202010639947 A CN 202010639947A CN 111757258 A CN111757258 A CN 111757258A
- Authority
- CN
- China
- Prior art keywords
- partition
- target
- model
- probability
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000010276 construction Methods 0.000 title claims abstract description 14
- 238000005192 partition Methods 0.000 claims abstract description 110
- 239000000203 mixture Substances 0.000 claims abstract description 28
- 238000000034 method Methods 0.000 claims abstract description 20
- 230000008878 coupling Effects 0.000 claims abstract description 11
- 238000010168 coupling process Methods 0.000 claims abstract description 11
- 238000005859 coupling reaction Methods 0.000 claims abstract description 11
- 238000012706 support-vector machine Methods 0.000 claims abstract description 6
- 238000012360 testing method Methods 0.000 claims description 24
- 238000005259 measurement Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 11
- 239000000470 constituent Substances 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 4
- 238000013145 classification model Methods 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 238000004088 simulation Methods 0.000 claims 1
- 230000008859 change Effects 0.000 abstract description 5
- 230000009467 reduction Effects 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 16
- 230000000875 corresponding effect Effects 0.000 description 15
- 238000009825 accumulation Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000000638 solvent extraction Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/021—Services related to particular areas, e.g. point of interest [POI] services, venue services or geofences
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/023—Services making use of location information using mutual or relative location information between multiple location based services [LBS] targets or of distance thresholds
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/33—Services specially adapted for particular environments, situations or purposes for indoor environments, e.g. buildings
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Remote Sensing (AREA)
- Image Analysis (AREA)
Abstract
A self-adaptive positioning fingerprint database construction method under a complex indoor signal environment belongs to the technical field of indoor positioning. According to the AP laying position and the space structure, the system adopts a one-to-many support vector machine algorithm to perform partition operation on a target area so as to accurately determine the area range of signal change. A multivariate Gaussian mixture model based on mutual interference between signals is established by using the coupling relation between the signals in the narrow and small subareas so as to improve the reduction of the positioning precision caused by signal fluctuation. When the indoor environment changes, the self-adaptive updating algorithm based on the partition multivariate Gaussian mixture model can judge the credibility of fingerprint data of each partition, and updates the model parameters of the partitions with larger signal fluctuation by the self-adaptive algorithm, so that the coupling degree between the model and the existing environment is improved. The experimental result shows that the method can utilize relatively small amount of sample data to construct a stable and maintainable indoor signal distribution model, and compared with other algorithms, the positioning accuracy is improved to a certain extent.
Description
Technical Field
The invention relates to an offline fingerprint database construction optimization technology for indoor positioning, and belongs to the technical field of indoor positioning.
Background
Global satellite navigation systems have been widely used to provide location services to people in outdoor environments, but the lack of signals also makes the systems unable to function in complex indoor environments. Due to the wide spread of WiFi facilities and the popularization of smart phones, indoor positioning systems based on Received Signal Strength (RSS) values are receiving close attention of a large number of researchers. However, since the design of wireless communication facilities is not intended to provide indoor navigation for people, how to reduce the positioning influence caused by uncertain interference of environment fluctuation on wireless signals becomes a difficult point which has to be faced and solved by the existing research.
In a positioning system based on existing wireless facilities, common commercial devices of the positioning system do not have an autonomous editable function and only can provide indoor general RSS measurements, so that the conventional methods based on time difference of arrival and distance difference of arrival cannot be directly translated and applied. By utilizing the fingerprint positioning algorithm of matching operation between the measurement signal and the actual position, the loss of the signal sent by the wireless facility on the time and space characteristics is made up, the mapping of the signal environment and the actual scene can be effectively realized, and the indoor positioning based on the RSS value is possible.
Disclosure of Invention
For a large indoor scene, the method consumes a large amount of manpower and material resources and is easily influenced by environmental factors, the mapping relation between the constructed offline fingerprint database and the signal distribution in the actual scene is weakened due to time change, and the offline fingerprint database needs to be continuously corrected so as to reduce the accumulation of mapping errors caused by time accumulation.
The technical scheme of the invention is as follows:
a self-adaptive positioning fingerprint database construction method under a complex indoor signal environment comprises three stages: the method comprises the following steps of an off-line stage, an on-line positioning stage and an on-line fingerprint database updating stage, wherein the specific steps are as follows:
step one, dividing target area grids and constructing an offline fingerprint database
The target region is divided into K regions omegakK ∈ {1, 2.., K } component, which is the region ΩkDivision into NkEach grid, the geometric center of each grid is taken as a reference pointWhere N ∈ {1, 2., Nk},Is 2 × NkThe dimensional position matrix represents the RP position. For each RP locationThe corresponding region is marked asWhereinAnd isi ≠ k, indicating that the reference point is located in the k partition. In thatCollected from MkRSS sample value of APWhereinIs shown inThe collected RSS sample value from the mth AP, M ∈ {1,2k}。
Step two, constructing SVM partition model of online positioning stage
And (3) setting a probability classification model of the support vector machine in a one-to-many mode, carrying out two-classification identification on each preset partition according to whether a target is located in the partition, and constructing an SVM model of each partition through training data. For given K partitions, K SVM models are set, and measurement signals of all reference APs in each partition are taken to form current observation data r ═ r1,r2,...,rM]Where M is the number of APs in the target received area, and-100 db is taken for the unreceived signal values. Aiming at the current observation data, each partitioned SVM model gives the distribution probability p (y) of whether the target is positioned in the corresponding region k1| r), wherein ykIdentify for partition, indicate that the target is located in partition ΩkInner, K ∈ {1, 2.., K }. The distribution probability p (y) given by each partitioned SVM modelk1 r), making a preliminary judgment on the partition where the target is located, and using the preliminary judgment as a primary judgment basis.
And (3) adopting partition operation based on a probability SVM, dividing the reference point observation data obtained in the step one into a training set and a test set, and training a partition judgment model SVM. The probability value of the target in the corresponding partition is obtained through the K SVM models, and the problem of misjudgment of the test point near the partition handover is solved by setting a secondary judgment basis. Selecting 2 subarea areas with the maximum probability in the judged area, namely p (y)i1| r) and p (y)j1| r), i, j ∈ {1,2i=1|r)>p(y j1| r), the difference being expressed as:
Δyp=p(yi=1|r)-p(yj=1|r), (1)
when Δ ypWhen the value is more than delta y, the influence of the partition i on the test point is far larger than that of the partition j, and the reference point is determined to be the partition i, wherein the value delta y is a secondary determination threshold value. For Δ ypIf the difference is less than delta y, both the areas are judged as target areas, corresponding area matching operation is respectively carried out, and the target positions obtained by respective subareas are subjected to probability averaging to obtain the final position estimation
Step three, constructing a multivariate Gaussian mixture model (MVGMM)
The method comprises the steps of establishing a multi-element Gaussian mixture model MVGMM by utilizing the correlation of different AP signals, weighting and approximately simulating the joint distribution condition between the RP position in a partition and each obtained AP signal by utilizing the probability density function of different parameters by continuously increasing the number of Gaussian elements, and making up for the neglect of the coupling relation between the AP signals in the traditional fingerprint database construction work. The probability distribution function of the multivariate gaussian mixture model is expressed as:
wherein C represents the number of constituent elements, pN(x|μc,Pc) Represents the mean value of μcCovariance of PcOf the weight wcThe sum of (a) and (b) is 1.
Based on equation (2), using partition omegakThe posterior probability of the inner RP location and RSS signal value joint distribution represents the multivariate gaussian mixture model as:
wherein, y k1 indicates that the target is in the kth partition, r indicates the RSS value of each AP signal received at reference point x,is the weight of the constituent elements of the multivariate Gaussian mixture model,is the average value of the elements and is the average value of the elements,is the element covariance, CkIs the number of constituent elements.
The MVGMM model adopts an EM algorithm to estimate model parameters. MVGMM model selection CkMean value ofCovariance ofFitting the partition omegakInner NkA reference pointWith the RSS value of each AP signal acquired in the corresponding partitionThe joint probability distribution of (c). Clustering sample data into C by using k-means algorithmkAnd (4) taking the mean value and the covariance of each cluster to initialize the EM algorithm parameters, wherein the initial weight setting of each cluster is the same. The log-likelihood form of the joint probability distribution is expressed as:
wherein z isk=[xk;Rk],Hidden variable gammac,nRepresenting the probability that the nth sample value belongs to the c-th gaussian component element.
Thereby determining observation dataThe probability of belonging to the c-th gaussian component element is expressed as:
calculating the weight of a multivariate Gaussian function by equation (5)Mean valueSum covarianceRespectively as follows:
for different CkThe value repeat clustering and EM estimation process offline fingerprint library is represented as (w)k;μk;Pk) Where K ∈ {1, 2., K }.
Fourthly, positioning the target in the domain
And estimating the target position in the area through an off-line fingerprint database. If the target at the current moment accepts the RSS values of M APs asObtaining the distribution probability p (y) of the target in each subarea through a subarea judgment modelk=1|rnow),k∈{1,2,...,K}。
Using the current measured value rnowSelecting corresponding sub-region omegakAP signal measurement value effective for internal target positioningObtaining the posterior probability distribution of the target position under given observation data according to the conditional probability criterion of the multivariate Gaussian distribution:
based on the resulting off-line fingerprint library (w)k;μk;Pk) Obtaining:
combining the partition probability with the formula (9) to obtain the target in the partition omegakThe posterior distribution probability of each internal reference point is as follows:
target is located in partition ΩkThe distribution weight of each reference point in the image is updated as follows:
wherein, λ is normalization factor, then the division is ΩkThe distribution weight of the internal reference points as the target positions is
In summary, the position of the target is estimated as
Step five, updating the fingerprint database
When dividing into omegakWhen the entropy of the obtained information is less than the partition threshold value, newly collecting NkData of a personAnd updating parameters of the partition model. Based on the existing multivariate Gaussian mixture model, model parameters are adaptively adjusted through data acquisition, so that an MVGMM model which has a closer coupling relation with the existing environment is deduced. In line with the EM algorithm, the adaptation process of the MVGMM model parameters is also a two-step estimation. Firstly, newly-added observation data is added through a k-means algorithmCluster as CkClustering, and obtaining data weight by clusteringMean valueCovarianceThe gaussian components are initialized. Respectively calculating the statistical weight of the newly added observation data by using the formulas (5) to (8)Mean valueAnd covariance
Dividing omega by using statistic parameter of new datakThe corresponding MVGMM model parameters are subjected to self-adaptive optimization, namely:
wherein,rho ∈ { w, m, v } is used for balancing influence degree of new and old data on model parameters, lambda is a normalization factor, rρIs an intrinsic correlation factor of the parameter.
Sixthly, updating the model parameters
And (5) carrying out real-time updating on the model parameters obtained by updating in the self-adaptive process in the second step and the third step.
The invention has the beneficial effects that: according to the AP laying position and the space structure, the system adopts a one-to-many support vector machine algorithm to perform partition operation on a target area so as to accurately determine the area range of signal change. A multivariate Gaussian mixture model based on mutual interference between signals is established by using the coupling relation between the signals in the narrow and small subareas so as to improve the reduction of the positioning precision caused by signal fluctuation. When the indoor environment changes, the self-adaptive updating algorithm based on the partition multivariate Gaussian mixture model can judge the credibility of fingerprint data of each partition, and updates the model parameters of the partitions with larger signal fluctuation by the self-adaptive algorithm, so that the coupling degree between the model and the existing environment is improved.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is an experimental scene diagram.
Fig. 3 is a comparison graph of the partitioning effect, in which (a) is the case of area partitioning and (b) is the diagram of the effect of area partitioning after training.
Fig. 4 is a comparison graph of RSS fingerprint map construction effect, in which (a) is true measurement, (b) is GP type estimation, and (c) is MVGMM model estimation.
Fig. 5 is a graph comparing the fit of the AP3 signal within partition one.
Fig. 6 is a comparison diagram of the trajectory prediction of the target motion.
FIG. 7 is a plot of a trajectory estimation error box.
FIG. 8 is a graph comparing error accumulation functions.
Figure 9 is a comparison of the effect of the AP3 data fit before and after a fingerprint library update. Wherein, (a) is the AP3 signal strength attenuation value, and (b) is the AP3 signal prediction error.
Detailed Description
Aiming at the problems of large data volume, high maintenance cost and the like of a large indoor scene, the method accurately maintains the region through partition operation and provides a partition Multivariate Gaussian Mixture Model (MVGMM) according to the coupling relation between signals in the partition so as to improve the fitting degree of signal distribution. The model divides a target area according to the position of a signal Access Point (AP) and a physical communication structure, and realizes partition operation through a one-to-many support vector machine model. In a relatively narrow subarea area, a multivariate Gaussian mixture model is respectively established by utilizing mutual interference existing between signals so as to strengthen the fitting degree of the signals and finally achieve the effect of improving the subarea positioning precision. When the environment changes, the algorithm takes the information entropy as a partition data updating criterion, and the influence of partition change on the fingerprint database is responded in time, so that the maintenance cost is reduced. Therefore, in the indoor positioning application, the construction of a fingerprint library which is supported by a small amount of data and can be maintained efficiently is realized.
The target positioning system based on the multivariate Gaussian mixture model can accurately position the target at lower acquisition cost. The algorithm takes the measurement signals acquired in the traditional WiFi facility as training data to construct the distribution condition of AP signals in the target area. Compared with the traditional target positioning process, the system consists of three modules, including an off-line stage, an on-line positioning stage and an on-line fingerprint database updating stage, as shown in fig. 1.
In the off-line stage, a complete fingerprint database is established by locating the positions of Reference Points (RP) in the area and the received RSS signal values, and is applied to the matching operation in the on-line stage. In the on-line positioning stage, a partition model is adopted to determine the area of the target, the conditional probability criterion of a multivariate Gaussian function is utilized to calculate the posterior probability of the target position with the current observation data as the condition, and the target position is estimated through WKNN. And in the online fingerprint database updating stage, the information entropy obtained by posterior distribution probability calculation is used as the updated measurement scale of the fingerprint database, and a feedback scheduling fingerprint database based on a partition multivariate Gaussian mixture model is established.
The method comprises the following specific steps:
constructing an offline fingerprint library
Fingerprint collection scheme
The target region is divided into K regions omegakK ∈ {1, 2.., K } and will be the region ΩkDivision into NkEach grid, the geometric center of each grid is taken as a reference pointWhere N ∈ {1, 2., Nk},Is 2 × NkThe dimensional position matrix represents the RP position. For each RP locationThe corresponding region is marked asWhereinAnd isi ≠ k, indicating that the reference point is located in the k partition. In thatCollected from MkRSS sample value of APWhereinIs shown inThe collected RSS sample value from the mth AP, M ∈ {1,2k}。
Partition model construction
By comprehensively considering the accuracy and efficiency of regional classification, the problem of classification of partitions set according to the communication between AP location and physics can be effectively solved by setting up a probability classification model of a support vector machine in a one-to-many manner [14 ]]. For each preset partition, whether the target is the target or notAnd performing two-classification identification on the subareas, and constructing an SVM model of each subarea through training data. For given K partitions, K SVM models are set, and measurement signals of all reference APs in each partition are taken to form current observation data r ═ r1,r2,...,rM]Where M is the number of APs in the target received area, and-100 db is taken for the unreceived signal values. For the current observation data, each partitioned SVM model can give the distribution probability p (y) of whether the target is located in the corresponding region k1| r), wherein ykIdentify for partition, indicate that the target is located in partition ΩkInner, K ∈ {1, 2.., K }. The distribution probability p (y) given by each partitioned SVM modelk1 r), the primary judgment can be made on the partition where the target is located, and the primary judgment can be made according to the partition where the target is located
The algorithm adopts partition operation based on a probability SVM, divides reference point observation data obtained in an offline stage into a training set and a test set, and trains a partition judgment model. The probability value of the target in the corresponding subarea can be obtained through the K SVM models, but the signal distribution at the subarea junction is complex, and judgment errors of the subarea models are easily caused. Therefore, the problem of misjudgment of the test points near the partition handover is solved by setting a secondary judgment basis. Selecting 2 subarea areas with the maximum probability in the judged area, namely p (y)i1| r) and p (y)j1| r), i, j ∈ {1,2i=1|r)>p(y j1| r), the difference can be expressed as
Δyp=p(yi=1|r)-p(yj=1|r), (1)
When Δ ypWhen the value is more than delta y, the influence of the partition i on the test point is far larger than that of the partition j, and the reference point can be determined in the partition i, wherein the delta y is a secondary determination threshold value. For Δ ypIf < delta y, judging both areas as target areas, respectively performing corresponding area matching operation, and performing probability average on target positions obtained by respective partitions to obtain final position estimation
Construction of multivariate Gaussian mixture model
The radiation pattern of the antenna is not directionally uniform due to attenuation and reflection of electromagnetic signals by walls in a small area. The signal strength values of different APs at a certain time are also correlated due to wall refraction, personnel occlusion, channel problems, etc. Thus, simply passing a single AP sample value splits the correlation between AP signals, resulting in large errors in model construction.
Considering the mutual interference among signals in the partitioned narrow and small subareas, a multivariate Gaussian mixture model can be established by utilizing the correlation of different AP signals, and the neglect of the coupling relation among the AP signals in common work is compensated by utilizing the probability density function weighting of different parameters and the combined probability density distribution between the RP position in the approximate subarea and the obtained AP signals through continuously increasing the number of Gaussian elements. The probability distribution function of the multivariate Gaussian mixture model can be expressed as
Wherein C represents the number of constituent elements, pN(x|μc,Pc) Represents the mean value of μcCovariance of PcOf the weight wcThe sum of (a) and (b) is 1.
Based on equation (2), using partition omegakThe posterior probability of the joint distribution of inner RP positions and RSS signal values can represent a multivariate Gaussian mixture model as
Wherein, y k1 indicates that the target is in the kth partition, r indicates the RSS value of each AP signal received at reference point x,is the weight of the constituent elements of the multivariate Gaussian mixture model,is the average value of the elements and is the average value of the elements,is the element covariance, CkIs the number of constituent elements.
In order to improve the fitting effect of the MVGMM model on the sample data acquired in the target area, the model parameters are estimated by adopting an EM algorithm. MVGMM model selection CkMean value ofCovariance ofFitting the partition omegakInner NkA reference pointAnd the obtained RSS value of each AP signalThe joint probability distribution of (c). Clustering sample data into C by using k-means algorithmkAnd (4) taking the mean value and the covariance of each cluster to initialize the EM algorithm parameters, wherein the initial weight setting of each cluster is the same. The log-likelihood form of the joint probability distribution can be expressed as
Wherein z isk=[xk;Rk],Hidden variable gammac,nRepresenting the probability that the nth sample value belongs to the c-th gaussian component element.
Thereby determining observation dataThe probability of belonging to the c-th Gaussian component element can be expressed as
The weight of the multivariate Gaussian function can be calculated by the formula (5)Mean valueSum covarianceRespectively as follows:
for different CkValue iterative clustering and EM estimation process:
(1) inputting: number of Gaussian composition elements CkInitial weight of Gaussian componentInitial mean valueInitial covarianceSection omegakLocation of internal reference pointAnd corresponding sampled valuec∈{1,2,..,Ck}
Based on the above analysis, the offline fingerprint library can be represented as (w)k;μk;Pk) Where K ∈ {1, 2., K }.
An online stage: target localization
And constructing an offline fingerprint database about the target area by using the acquired sample data, and estimating the target position in the area by using the fingerprint database. If the target at the current moment accepts the RSS values of M APs asThe distribution probability p (y) of the target in each partition can be obtained through the partition judgment modelk=1|rnow),k∈{1,2,...,K}。
Using the current measured value rnowCorresponding sub-region omega can be selectedkAP signal measurement value effective for internal target positioningConditional probability criterion based on multivariate Gaussian distributionThen, the posterior probability distribution of the target location under the given observation data can be obtained
Wherein the off-line fingerprint library (w) is obtained based onk;μk;Pk) Is obtained by
Combining the partition probability with equation (9), the target can be obtained in partition omegakThe posterior distribution probability of each internal reference point is
Target is located in partition ΩkThe distribution weight of each reference point in the table can be updated to
Wherein, λ is normalization factor, then the division is ΩkThe distribution weight of the internal reference points as the target positions is
In summary, the position of the target is estimated as
An online stage: fingerprint library update
When dividing into omegakWhen the entropy of the obtained information is less than the partition threshold value, newly collecting NkData of a personAnd updating parameters of the partition model. Based on the existing multivariate Gaussian mixture model, model parameters can be adaptively adjusted through data acquisition so as to deduce an MVGMM model which has a closer coupling relation with the existing environment. Consistent with the EM algorithm, the self-adaption process of the MVGMM model parameters is also a two-step estimation[13]. Firstly, newly-added observation data is added through a k-means algorithmCluster as CkClustering, and obtaining data weight by clusteringMean valueCovarianceThe gaussian components are initialized. The statistical weights of the newly added observation data can be respectively calculated by using the formulas (5) to (8)Mean valueAnd covariance
Dividing omega by using statistic parameter of new datakThe corresponding MVGMM model parameters are subjected to self-adaptive optimization, namely:
wherein,rho ∈ { w, m, v } is used for balancing influence degree of new and old data on model parameters, lambda is a normalization factor, rρIs an intrinsic correlation factor of the parameter. The newly added data can better reflect the distribution condition of the signals in the current environment, and in the self-adaptive updating process, the balance factor is shown as followsρ ∈ { w, m, v } is more dependent on the newly added observation data.
(1) Inputting: original Gaussian component number CkWeight of original Gaussian componentMean valueCovarianceNewly-added partition omegakInternally sampled datac∈{1,2,..,Ck};
Illustrating according to what is contained in the claims
Example 1: personnel position detection
The position information of indoor personnel is detected in real time, each target personnel is required to be equipped with an intelligent bracelet, and the target position is determined through AP signals received by the intelligent bracelet. But since the indoor environment is complicated and the AP device is changed by time. The SMVGMM algorithm is adopted to fit the distribution condition of the AP signals in the room, the positioning precision can be improved, the influence of environmental factors can be reduced, and the model parameters can be updated through the adaptive updating algorithm in the online updating stage, so that the influence of time change on an offline fingerprint library can be reduced.
The test scene is an annular corridor environment of a certain layer in a C area of a certain college, and WiFi routers uniformly laid by mobile operators in the college are selected as AP signal sources. Because the AP signal source is mainly laid in the corridor, and the unilateral corridor area is relatively wide, and the difference that the signal is influenced by the wall is less, so relatively divide into corresponding subregion together with and receive the less unilateral corridor area of AP signal difference with physical structure, then experimental region can divide into K4 subregion. The RP positions adopt a mesh topology and are arranged in a corridor wide-width centered mode, adjacent RPs are spaced by 1 meter, 368 RP points are counted, and an arrangement plan view of the AP signal source and the RP points is shown in FIG. 2. And according to the stability of the AP signal sources in each partition, the number of the AP signal sources in the selected areas 1-4 is {4,5,4,4 }. In order to reduce the influence of equipment difference on a positioning algorithm, a unified model smart phone is used for signal collection in an experiment.
And acquiring signal intensity values of 12 APs selected by all the partitions at all the reference points, wherein the sampling interval is 1.2s, the sampling interval is 4.8s (data caching caused by frequency reasons of equipment is avoided), and constructing an offline fingerprint library according to the process in the section 3.1. In the testing stage, an experimenter holds the same-style smart phone to walk for a circle along a testing area, the same-style smart phone is advanced to a testing point to obtain real-time observation data through operation, the current position is marked, 184 testing points are obtained in the testing process, and the interval is 1 meter.
After the target area is divided, the acquisition reference points may be identified by partitions, and as shown in fig. 3 (a), a partition model is constructed by sampling data and the partition identifications. Based on the partition model, the test data can be partitioned, the partition effect is shown as (b) in fig. 3, and the algorithm divides the test points needing to start the second-level criterion into an area 5 to represent a signal complex area.
In order to verify the construction effect of the algorithm on the RSS fingerprint map, a true real number map of an AP3 signal in a target area and an estimated effect map of an MVGMM model and a GP model are shown in FIG. 4. After fitting the joint distribution of the reference point location and the AP signal in the target region by using the MVGMM model, the RSS distribution of the AP3 signal in the region can be obtained by adjusting the corresponding location using equations (7) and (8). The RSS measurement value of the AP3 signal is extracted from the observation data of the reference point in the test, 50% of test value is selected as training data, the model is trained, and the RSS distribution effect is verified by using the remaining 50% of measurement value. Since different partitions are selected without using AP facilities and AP3 mainly acts in partition one, FIG. 5 shows the fitting effect of two models on AP3 signals at part of the reference points in partition one. By comparing the actual measurement value with the model estimation value, it can be seen that the MVGMM model better fits the distribution condition of the AP signal in the indoor environment through a plurality of gaussian component elements, and is particularly embodied in partition one.
Based on the obtained MVGMM model and the test data, the patent (abbreviated as SMVGMM) is compared with the traditional WKNN algorithm and GP algorithm respectively, and the positioning accuracy of the algorithm is analyzed. The user makes one turn in the target area and the comparison of the position estimates for the three algorithms is shown in fig. 6. As can be seen from fig. 6, the target travel track prediction obtained in the patent is smoother, and the overall positioning accuracy is improved compared to the GP algorithm. FIG. 7 shows a plot of the error box for the three algorithms at each test point. As can be seen from the figure, compared with the GP algorithm, the whole-course positioning accuracy of the patent is improved by more than 20%, and the smoothness of the target predicted track obtained by the patent is verified in the near step.
Due to the great difference between the WKNN algorithm and the other two algorithms in the positioning accuracy, fig. 8 only shows the cumulative probability comparison graph of the position estimation errors of the patent and GP algorithms. As can be seen from the figure, the initial error accumulation speed of the patent is slower than that of the GP algorithm, the overall effect is better than that of the GP algorithm, and the positioning effect of the traditional algorithm is comprehensively improved by the correlation between AP signals in narrow subareas of the patent.
In order to verify the effect and value of online updating of the fingerprint database, the experiment carries out data acquisition on a target area twice (at a one-week interval), and carries out self-adaptive updating on the MVGMM model parameters constructed by the original data by using the acquired data for the second time. And comparing the fitting effect of the MVGMM model on the AP3 signal before and after parameter updating through the test data acquired twice. Fig. 9 shows the fitting condition of the model to the AP3 signal before and after the parameter update, and the fitting error thereof. It can be seen from the figure that the test data acquired twice have a large difference in the region four, and the fitting effect of the model after parameter updating on the latest test data is superior to that of the previous model, and is particularly embodied in the region four.
Claims (1)
1. A self-adaptive positioning fingerprint database construction method under a complex indoor signal environment is characterized by comprising three stages: the method comprises the following steps of an off-line stage, an on-line positioning stage and an on-line fingerprint database updating stage, wherein the specific steps are as follows:
step one, dividing target area grids and constructing an offline fingerprint database
The target region is divided into K regions omegakK ∈ {1, 2.., K } component, which is the region ΩkDivision into NkEach grid, the geometric center of each grid is taken as a reference pointWhere N ∈ {1, 2., Nk},Is 2 × NkThe dimension position matrix represents the RP position; for each RP locationThe corresponding region is marked asWhereinAnd isi ≠ k, meaning that the reference point is located in the k partition; in thatCollected from MkRSS sample value of APWhereinIs shown inThe collected RSS sample value from the mth AP, M ∈ {1,2k};
Step two, constructing SVM partition model of online positioning stage
Setting a probability classification model of a support vector machine in a one-to-many mode, carrying out two-classification identification on each preset partition according to whether a target is located in the partition, and constructing an SVM model of each partition through training data; for given K partitions, K SVM models are set, and measurement signals of all reference APs in each partition are taken to form current observation data r ═ r1,r2,...,rM]Wherein M is the number of APs in the target received area, and the value of the unreceived signal is-100 db; aiming at the current observation data, each partitioned SVM model gives the distribution probability p (y) of whether the target is positioned in the corresponding regionk1| r), wherein ykIdentify for partition, indicate that the target is located in partition ΩkInner, K ∈ {1, 2.. multidot.K }, distribution probability p (y) given by each partitioned SVM modelk1 r), performing primary judgment on a partition where the target is located, and taking the partition as a primary judgment basis;
dividing the reference point observation data obtained in the step one into a training set and a test set by adopting partition operation based on a probability SVM, and training a partition judgment model SVM; the probability value of the target in the corresponding partition is obtained through K SVM models, and the problem of misjudgment of the test point near the partition handover is solved by setting a secondary judgment basis; selecting 2 subarea areas with the maximum probability in the judged area, namely p (y)i1| r) and p (y)j1| r), i, j ∈ {1,2i=1|r)>p(yj1| r), the difference being expressed as:
Δyp=p(yi=1|r)-p(yj=1|r), (1)
when Δ ypWhen the value is more than delta y, the influence of the partition i on the test point is far larger than that of the partition j, and the reference point is determined in the partition i, wherein the delta y is a secondary determination threshold value; for Δ ypIf the difference is less than delta y, both the areas are judged as target areas, corresponding area matching operation is respectively carried out, and the target positions obtained by respective subareas are subjected to probability averaging to obtain the final position estimation
Step three, constructing a multivariate Gaussian mixture model (MVGMM)
Establishing a multivariate Gaussian mixture model (MVGMM) by utilizing the correlation of different AP signals, and compensating the neglect of the coupling relation between the AP signals in the traditional fingerprint library construction work by continuously increasing the number of Gaussian elements and utilizing the probability density function weighting and approximate simulation of different parameters to simulate the joint distribution condition between the RP position in the subarea and the obtained AP signals; the probability distribution function of the multivariate gaussian mixture model is expressed as:
wherein C represents the number of constituent elements, pN(x|μc,Pc) Represents the mean value of μcCovariance of PcOf the weight wcThe sum of (a) and (b) is 1;
based on equation (2), using partition omegakThe posterior probability of the inner RP location and RSS signal value joint distribution represents the multivariate gaussian mixture model as:
wherein, yk1 indicates that the target is in the kth partition, r indicates the RSS value of each AP signal received at reference point x, is the weight of the constituent elements of the multivariate Gaussian mixture model,is the average value of the elements and is the average value of the elements,is the element covariance, CkIs the number of constituent elements;
the MVGMM adopts an EM algorithm to estimate model parameters; MVGMM model selection CkMean value ofCovariance ofFitting the partition omegakInner NkA reference pointWith the RSS value of each AP signal acquired in the corresponding partitionA joint probability distribution of (a); clustering sample data into C by using k-means algorithmkTaking the mean value and covariance of each cluster to initialize EM algorithm parameters, wherein the initial weight settings of each cluster are the same; the log-likelihood form of the joint probability distribution is expressed as:
wherein z isk=[xk;Rk],Hidden variable gammac,nRepresenting the probability that the nth sample value belongs to the c-th gaussian component element;
thereby determining observation dataThe probability of belonging to the c-th gaussian component element is expressed as:
calculating the height of the multiple element by the formula (5)Weight of a gaussian functionMean valueSum covarianceRespectively as follows:
for different CkThe value repeat clustering and EM estimation process offline fingerprint library is represented as (w)k;μk;Pk) K ∈ {1, 2.., K }, and step four, positioning the target in the subarea
Estimating the target position in the area through an offline fingerprint database; if the target at the current moment accepts the RSS values of M APs asObtaining the distribution probability p (y) of the target in each subarea through a subarea judgment modelk=1|rnow),k∈{1,2,...,K};
Using the current measured value rnowSelecting corresponding sub-region omegakAP signal measurement value effective for internal target positioningObtaining the posterior probability distribution of the target position under given observation data according to the conditional probability criterion of the multivariate Gaussian distribution:
based on the resulting off-line fingerprint library (w)k;μk;Pk) Obtaining:
combining the partition probability with the formula (9) to obtain the target in the partition omegakThe posterior distribution probability of each internal reference point is as follows:
target is located in partition ΩkThe distribution weight of each reference point in the image is updated as follows:
wherein, λ is normalization factor, then the division is ΩkThe distribution weight of the internal reference points as the target positions is
In summary, the position of the target is estimated as
Step five, updating the fingerprint database
When dividing into omegakWhen the entropy of the obtained information is less than the partition threshold value, newly collecting NkData of a personUpdating parameters of the partition model; based on the existing multivariate Gaussian mixture model, model parameters are adaptively adjusted through data acquisition so as to deduce an MVGMM model which has a closer coupling relation with the existing environment; consistent with the EM algorithm, the self-adaptive process of the MVGMM model parameters is also a two-step estimation; firstly, newly-added observation data is added through a k-means algorithmCluster as CkClustering, and obtaining data weight by clusteringMean valueCovarianceInitializing each Gaussian composition; respectively calculating the statistical weight of the newly added observation data by using the formulas (5) to (8)Mean valueAnd covariance
Dividing omega by using statistic parameter of new datakThe corresponding MVGMM model parameters are subjected to self-adaptive optimization, namely:
wherein,used for balancing the influence degree of new and old data on model parameters, wherein lambda is a normalization factor, and rρIs an intrinsic correlation factor of the parameter;
sixthly, updating the model parameters
And (5) carrying out real-time updating on the model parameters obtained by updating in the self-adaptive process in the second step and the third step.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010639947.2A CN111757258B (en) | 2020-07-06 | 2020-07-06 | Self-adaptive positioning fingerprint database construction method under complex indoor signal environment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010639947.2A CN111757258B (en) | 2020-07-06 | 2020-07-06 | Self-adaptive positioning fingerprint database construction method under complex indoor signal environment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111757258A true CN111757258A (en) | 2020-10-09 |
CN111757258B CN111757258B (en) | 2021-05-14 |
Family
ID=72679412
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010639947.2A Active CN111757258B (en) | 2020-07-06 | 2020-07-06 | Self-adaptive positioning fingerprint database construction method under complex indoor signal environment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111757258B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112533137A (en) * | 2020-11-26 | 2021-03-19 | 北京爱笔科技有限公司 | Device positioning method and device, electronic device and computer storage medium |
CN113518308A (en) * | 2021-05-25 | 2021-10-19 | 江南大学 | Optimal AP screening method in indoor positioning |
CN114048783A (en) * | 2021-11-17 | 2022-02-15 | 东南大学 | Cellular signal map construction method based on mobile group perception |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107087256A (en) * | 2017-03-17 | 2017-08-22 | 上海斐讯数据通信技术有限公司 | A kind of fingerprint cluster method and device based on WiFi indoor positionings |
CN107241700A (en) * | 2017-04-23 | 2017-10-10 | 西安电子科技大学 | Indoor orientation method based on CSI Space-frequency characters and reference point locations clustering algorithm |
CN109413575A (en) * | 2018-12-04 | 2019-03-01 | 重庆邮电大学 | A kind of adaptive AP layout method of indoor positioning |
US10571546B2 (en) * | 2015-09-25 | 2020-02-25 | Intel Corporation | Vision and radio fusion based precise indoor localization |
CN111148030A (en) * | 2019-12-31 | 2020-05-12 | 泰斗微电子科技有限公司 | Fingerprint database updating method and device, server and storage medium |
-
2020
- 2020-07-06 CN CN202010639947.2A patent/CN111757258B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10571546B2 (en) * | 2015-09-25 | 2020-02-25 | Intel Corporation | Vision and radio fusion based precise indoor localization |
CN107087256A (en) * | 2017-03-17 | 2017-08-22 | 上海斐讯数据通信技术有限公司 | A kind of fingerprint cluster method and device based on WiFi indoor positionings |
CN107241700A (en) * | 2017-04-23 | 2017-10-10 | 西安电子科技大学 | Indoor orientation method based on CSI Space-frequency characters and reference point locations clustering algorithm |
CN109413575A (en) * | 2018-12-04 | 2019-03-01 | 重庆邮电大学 | A kind of adaptive AP layout method of indoor positioning |
CN111148030A (en) * | 2019-12-31 | 2020-05-12 | 泰斗微电子科技有限公司 | Fingerprint database updating method and device, server and storage medium |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112533137A (en) * | 2020-11-26 | 2021-03-19 | 北京爱笔科技有限公司 | Device positioning method and device, electronic device and computer storage medium |
CN112533137B (en) * | 2020-11-26 | 2023-10-17 | 北京爱笔科技有限公司 | Positioning method and device of equipment, electronic equipment and computer storage medium |
CN113518308A (en) * | 2021-05-25 | 2021-10-19 | 江南大学 | Optimal AP screening method in indoor positioning |
CN113518308B (en) * | 2021-05-25 | 2022-03-25 | 江南大学 | Optimal AP screening method in indoor positioning |
CN114048783A (en) * | 2021-11-17 | 2022-02-15 | 东南大学 | Cellular signal map construction method based on mobile group perception |
CN114048783B (en) * | 2021-11-17 | 2024-04-16 | 东南大学 | Cellular signal map construction method based on mobile group perception |
Also Published As
Publication number | Publication date |
---|---|
CN111757258B (en) | 2021-05-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111757258B (en) | Self-adaptive positioning fingerprint database construction method under complex indoor signal environment | |
CN105491661B (en) | Indoor locating system and method based on improved Kalman filter algorithm | |
CN108174343B (en) | Wireless positioning method and system for power indoor communication operation and maintenance scene | |
Kaemarungsi | Efficient design of indoor positioning systems based on location fingerprinting | |
CN104038901B (en) | Indoor positioning method for reducing fingerprint data acquisition workload | |
US20050246334A1 (en) | Location determination and location tracking in wireless networks | |
CN108919177B (en) | Positioning map construction method based on virtual information source estimation and track correction | |
He et al. | INTRI: Contour-based trilateration for indoor fingerprint-based localization | |
Zhou et al. | Indoor fingerprint localization based on fuzzy c-means clustering | |
CN113574412A (en) | Offline adjustment system for detecting new motion zones in a motion detection system | |
CN109348416B (en) | Fingerprint indoor positioning method based on binary k-means | |
CN103533647A (en) | Radio frequency map self-adaption positioning method based on clustering mechanism and robust regression | |
CN103369571A (en) | Method for correction and coverage self-optimization of propagation model based on multi-network combined measurement | |
CN102253365B (en) | Indoor positioning method based on estimation of wireless signal source parameters | |
CN109814066A (en) | RSSI indoor positioning distance measuring method, indoor positioning platform based on neural network learning | |
CN101873605B (en) | Adaptive method for classifying communication environments in network planning | |
Nikitin et al. | Indoor localization accuracy estimation from fingerprint data | |
CN109143161B (en) | High-precision indoor positioning method based on mixed fingerprint quality evaluation model | |
CN105866732B (en) | The mixing indoor orientation method that a kind of improvement MK models and WKNN algorithms are combined | |
CN114449452A (en) | Indoor positioning algorithm for heterogeneous equipment | |
CN109121081B (en) | Indoor positioning method based on position candidate set and EM algorithm | |
CN108919182B (en) | Target positioning method based on support set and expectation maximization in WIFI environment | |
CN104821854B (en) | A kind of many primary user's multidimensional frequency spectrum sensing methods based on random set | |
CN109379702A (en) | A kind of three-dimension sensor network node positioning method and system | |
CN117241215A (en) | Wireless sensor network distributed node cooperative positioning method based on graph neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |