CN103987118B - Access point k means clustering methods based on received signal strength signal ZCA albefactions - Google Patents
Access point k means clustering methods based on received signal strength signal ZCA albefactions Download PDFInfo
- Publication number
- CN103987118B CN103987118B CN201410212396.6A CN201410212396A CN103987118B CN 103987118 B CN103987118 B CN 103987118B CN 201410212396 A CN201410212396 A CN 201410212396A CN 103987118 B CN103987118 B CN 103987118B
- Authority
- CN
- China
- Prior art keywords
- cluster
- received signal
- signal strength
- cluster centre
- access point
- 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.)
- Active
Links
Abstract
The invention discloses a kind of access point k means clustering methods based on received signal strength signal ZCA albefactions, each location fingerprint in RM is represented with the vectorial average of the received signal strength in corresponding reference point, and by mean normalization;Then by whitening received signal strength mean value, correlation is removed;K means clusters choose k fingerprint in whole RM and are used as initial cluster center;For other all received signal strength averages in addition to k cluster centre, then according to their Euclidean distances with these cluster centres, the cluster recently with its Euclidean distance is assigned these to respectively;Perform after all fingerprints, obtained new cluster, regard the average value of all fingerprints newly clustered as new cluster centre;Continuous repeat step three and four, until k cluster centre no longer changes, terminates iteration.The present invention substantially reduces the correlation between received signal strength signal, improves the degree of accuracy of cluster, so as to further improve the positioning precision of system.
Description
Technical field
The invention belongs to indoor positioning region clustering technical field, more particularly to it is a kind of for indoor positioning based on reception
The access point k-means clustering methods of signal strength signal ZCA albefactions.
Background technology
For larger positioning target area, the statistical property of received signal strength (RSS) changes greatly, for based on learning-oriented
For location algorithm, if learning to whole localization region, algorithm complex will be increased, the location model of foundation is not optimal
, so as to be unfavorable for the positioning precision of raising system.Therefore need to carry out clustering block to localization region, by larger positioning area
Domain is divided into several small localization regions, and models respectively, with the purpose for reaching reduction computation complexity, improving positioning precision.
Existing clustering algorithm does not all account for the relativity problem between received signal strength (RSS) signal, so as to cause classification essence
Degree is not high enough, limit the precision of position resolving.
Mainly there is following several method on localization region cluster:
1) the explicit clustering algorithm that Youssef et al. is proposed.RM location fingerprint is divided into different clusters by the algorithm, each
Cluster shares same access point (AP) set received, and this is collected to the key assignments for being collectively referred to as the cluster.This method is adapted to positioning
Target area less, in the case that access point (AP) quantity is few, target area and more access point is positioned for big
(AP) in the case of existing, often there is a situation where to sense in some reference points less than some access points (AP), this is for online
Positioning stage can make mobile terminal wrong because can not find correct cluster or can not resolve positional information.
2) Region Segmentation Algorithm that Borenovic et al. is proposed.The algorithm needs that localization region artificially is divided into face
Several suitable positioning subregions of product.The algorithm weak point is that localization region is divided into size one using artificial method
The subregion of cause, it is difficult to which the characteristic such as similitude of difference and signal with respect to indoor framework and layout, algorithm application is limited.
3) k averages (k-means) algorithm that Chen et al. is proposed.It is exactly not account for receiving signal in place of its technical deficiency
Relativity problem between intensity (RSS) signal, limits the degree of accuracy of cluster.
The content of the invention
The purpose of the embodiment of the present invention is to provide a kind of access point k- based on received signal strength signal ZCA albefactions
Means clustering methods, it is intended to solve due to correlation between received signal strength signal, the degree of accuracy of cluster is not high enough, limitation
Positioning precision problem.
The embodiment of the present invention is achieved in that a kind of access point k- based on received signal strength signal ZCA albefactions
Means clustering methods, being somebody's turn to do the access point k-means clustering methods based on received signal strength signal ZCA albefactions includes:
Step one, each location fingerprint in RM is represented with the vectorial average of the received signal strength in corresponding reference point,
And by mean normalization;Then by whitening received signal strength mean value, correlation is removed;
Step 2, k-means clusters choose k fingerprint in whole RM and are used as initial cluster center;
Step 3, for other all received signal strength averages in addition to k cluster centre, then according to them and this
The Euclidean distance of a little cluster centres, assigns these to the cluster recently with its Euclidean distance respectively;
Step 4, has been performed after all fingerprints, obtains new cluster, using the average value of all fingerprints newly clustered as
New cluster centre;
Step 5, continuous repeat step three and four, until k cluster centre no longer changes, terminates iteration.
Further, cluster terminates, and each location fingerprint is converged to cluster centre nearest therewith, just will each be clustered
It is considered as a positioning subregion;Off-line phase, each cluster and corresponding location fingerprint data constitute an independent sub- fingerprint
Database;Tuning on-line stage, the received signal strength newly measured by calculating the Euclidean distance with cluster centre, is obtained first
Nearest cluster centre, is then drawn the positioning subregion of user by the corresponding mapping function of this cluster centre.
The access point k-means clustering methods based on received signal strength signal ZCA albefactions that the present invention is provided, will be received
Signal intensity (RSS) signal first carries out albefaction, is then clustered in conjunction with k-means, can fully reduce received signal strength (RSS)
Correlation between signal, improves the degree of accuracy of cluster, so as to further improve the positioning precision of system.The method letter of the present invention
It is single, it is easy to operate, preferably solve due to correlation between received signal strength signal, the degree of accuracy of caused cluster is low,
The problem of positioning precision is low.
Brief description of the drawings
Fig. 1 is the access point k-means clusters provided in an embodiment of the present invention based on received signal strength signal ZCA albefactions
The flow chart of method;
Fig. 2 is the access point k-means clusters provided in an embodiment of the present invention based on received signal strength signal ZCA albefactions
The flow chart of the embodiment 1 of method;
Fig. 3 WLAN indoor positioning environmental structure figures;
Fig. 4 shows cluster centre number k from during 1 to 8 change, k-means clustering algorithms before and after RSS signal albefactions
Clustering precision compares figure;
Fig. 5 is system accuracy probability cumulative distribution figure when k takes different value;
Signal of the k-means clustering algorithms before and after RSS signals whether there is albefaction to the comparison of positioning precision when Fig. 6 is k=4
Figure;
Fig. 7 is maximum, minimum, the mean error of system position error before and after RSS signal albefactions, and position error standard
Variance statistic figure.
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
Below in conjunction with the accompanying drawings and specific embodiment to the present invention application principle be further described.
As shown in figure 1, the access point k-means based on received signal strength signal ZCA albefactions of the embodiment of the present invention gathers
Class method comprises the following steps:
S101:Each location fingerprint in RM is represented with the vectorial average of the received signal strength in corresponding reference point, and
By mean normalization.Then by whitening received signal strength mean value, correlation is removed;
S102:K-means clusters choose k fingerprint in whole RM and are used as initial cluster center;
S103:For other all received signal strength averages in addition to k cluster centre, then according to them and these
The Euclidean distance of cluster centre, assigns these to the cluster recently with its Euclidean distance respectively;
S104:Perform after all fingerprints, obtained new cluster, using the average value of all fingerprints newly clustered as new
Cluster centre;
S105:Continuous repeat step S103 and S104, until k cluster centre no longer changes, terminate iteration.
The present invention's concretely comprises the following steps:
1) each location fingerprint in RM is represented with the vectorial average of the received signal strength (RSS) in corresponding reference point,
And by mean normalization.Then by whitening received signal intensity (RSS) average, correlation is removed.
2) k-means clusters choose k fingerprint as initial cluster center (in order to avoid randomly selected in whole RM
Initial cluster center causes the concentrations or deployment conditions of cluster, and it is individual initial poly- equably to choose k in positioning subregion as far as possible
Class center, is clustered with realizing as far as possible according to the uniformity in physical location space).
3) for other all received signal strength (RSS) averages in addition to k cluster centre, then according to them and this
The Euclidean distance of a little cluster centres, assigns these to the cluster recently with its Euclidean distance respectively.
4) performed after all fingerprints, obtained new cluster, using the average value of all fingerprints newly clustered as new
Cluster centre.
5) continuous repeat step (3) and (4).Until k cluster centre no longer changes, iteration is terminated.
Cluster terminates, and each location fingerprint is converged to cluster centre nearest therewith, and each cluster just is considered as into one
Individual positioning subregion.Off-line phase, each cluster and corresponding location fingerprint data constitute an independent sub- fingerprint database.
Tuning on-line stage, the received signal strength newly measured (RSS) is obtained most by being calculated with the Euclidean distance of cluster centre first
Nearly cluster centre, is then drawn the positioning subregion of user by the corresponding mapping function of this cluster centre.
Embodiments of the invention:
The experimental situation of the present invention is that the typical office block 5 layers of a building is completed, fragment of brick wall, metal window, glass
Glass and timber.This building is covered by 802.11 WLANs, most visible 46 access points.Specific experiment region is building
The 4th layer, general 400 square metres of area.Whole floor has computer room laboratory, corridor, hall, a hourly halt
And a lavatory, and some offices.In this experiment, 68 reference points (RP) and 35 random test points are laid altogether
(TestPoint,TP).40 RSS samples are probably received in each RP, and each TP then receives general 5~20 RSS measurements
Value.All data acquisitions are all completed on same working day.The location algorithm of on-line stage uses WKNN (K=3).Experimental ring
Border is as shown in figure 3, red "+" therein is RP positions.
Fig. 4 show cluster centre number k from 1 to 8 change when, the k-means clustering algorithms before and after RSS signal albefactions
Clustering precision compare figure.
As can be seen from Figure 4 the k-means algorithms first after RSS albefactions cluster the degree of accuracy in terms of be better than albefaction it
Preceding effect.Especially with the increase of cluster centre k values, this raising is further obvious, because, in positioning target area
In the case of certain, cluster centre value is bigger, illustrates that the positioning subregion being divided is more, then signal is strong between corresponding neighborhood
Correlation between degree is bigger, and the redundancy that signal correlation is brought is removed by albefaction, so as to so that poly- before and after albefaction
The class piecemeal degree of accuracy has larger lifting.Secondly, in this experiment, when k value takes 4, the degree of accuracy of cluster can reach
To 97%, certainly, the accuracy rate clustered when k is less than 4 is higher, but when k values are too small, information in class easily occurs
Similarity is not high, is unfavorable for reduction computation complexity and improves positioning precision.In addition, from this figure it can be seen that with cluster
The increase of number, the piecemeal precision of the k-means clustering algorithms before and after all albefactions all drops at the fast speed, such as when k values are more than
After 5, due to excessive piecemeal so that each locator regional extent reduces, cause similarity increase between class, not only drop
The low piecemeal degree of accuracy, while can also reduce positioning precision.
It is the corresponding system position error probability cumulative distribution when k value gets 8 by 1 as shown in Figure 5.
From fig. 5, it can be seen that when k value is 4, the positioning precision of system takes the situation of other values better than k on the whole,
Although the cumulative probability at less than or equal to 3 meters is not highest, this has no effect on its overall positioning performance advantage.It is true
On, cumulative probability during k=4 is 77.1% herein, and maximum herein is 80% (k=5), although can when k values are 5
To realize within 3 meters of positioning precision 80% fiducial probability, but because clustering precision now only has 93.9%, hence it is evident that less than k
97% when taking 4, and positioning precision of cluster centre value when being 5 within 2 meters and within 4 meters be not high, so comprehensive
Consider to show of both the clustering block degree of accuracy and positioning precision, it is believed that most beneficial for system when cluster centre number is 4
Reliability of positioning.From Fig. 5, we can also be seen that cluster is conducive to improving positioning precision, such as when cluster centre number point
When not taking 2,3,4,5, the cumulative probability distribution of system accuracy is all better than the situation without cluster (k=1).But also see simultaneously
Arrive, when k value is 6,7 and 8, the good behaviour of this cluster is disappeared again, that is to say, that in putting for positioning precision
To be less than without cluster situation in terms of believing probability.
Above-mentioned analysis shows, the numerical value of cluster centre has direct relation for the performance quality of system, so in order to
Location fingerprint space can be reduced, positioning precision is improved, positioning subinterval can be effectively divided again, it is ensured that higher cluster is accurate
Property, it is to avoid the harmful effect because of cluster error band to positioning precision, cluster centre k values will be taken to be taken as 4 herein.It is 4 when taking k values
When, we compare comparison of the k-means clustering algorithms to positioning precision before and after RSS whether there is albefaction, as shown in Figure 6.
In the case of Fig. 6 optimum clusters, influence of the signal albefaction to positioning precision
From fig. 6, it can be seen that to being clustered again after signal albefaction, the positioning precision of system can be significantly improved.After albefaction
Probability of the cluster positioning precision in 2 meters be 60%, improve 39.7% compared with cluster positioning of the RSS without albefaction;In 3 meters
Positioning precision probability is 77.1%, and 12.4% is improved than the cluster positioning without albefaction.In addition, the positioning precision in 1 meter is general
Rate also has more obvious increase.In order to illustrate more clearly of the effect of albefaction, Fig. 7 compared for system position error before and after albefaction
Maximum, minimum, mean error, and position error standard variance.It can be seen from figure 7 that the system positioning after albefaction is flat
Equal error is 2.21 meters, and the positioning mean error without albefaction is 2.74 meters, improves 24%;And maximum positioning error is also by nothing
7.91 meters of albefaction drop to 6.27 meters, have dropped 1.64 meters;In addition, the standard variance of position error is also than 2.85 meters without albefaction
It is changed into 1.67 meters, this error illustrated in each test point deviates the degree of average localization error in reduction, from position error model
The change enclosed can also illustrate this point, and the error range before albefaction is 0.93~7.91 meter, the cluster position error scope of albefaction
For 0.61~6.27 meter.Because orientation range is smaller, the degree that specification error deviates average is just smaller, so standard variance just meeting
Diminish.
Embodiment 1:
As shown in Fig. 2 embodiments of the invention specifically include following steps:
The first step, it is determined that cluster number k, calculates the RSS averages from all reference points and normalize;
Second step, albefaction normalization data;
3rd step, initializes k cluster centre;
4th step, calculates the corresponding normalization RSS averages of RP and the distance of k cluster centre;
5th step, reference point is divided into the class where nearest cluster centre, suppresses to update each cluster centre;
Whether the 6th step, cluster centre immobilizes, no, then returns to the 4th step, be then to perform next step;
7th step, k cluster centre of output and correspondence RP set.
Received signal strength (RSS) signal is first carried out albefaction by the present invention, is then clustered in conjunction with k-means, can be abundant
Reduce the correlation between received signal strength (RSS) signal, improve the degree of accuracy of cluster, so as to further improve determining for system
Position precision.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
Any modifications, equivalent substitutions and improvements made within refreshing and principle etc., should be included in the scope of the protection.
Claims (2)
1. a kind of access point k-means clustering methods based on received signal strength signal ZCA albefactions, it is characterised in that the base
Include in the access point k-means clustering methods of received signal strength signal ZCA albefactions:
Step one, each location fingerprint in RM is represented with the vectorial average of the received signal strength in corresponding reference point, and will
Mean normalization;Then by whitening received signal strength mean value, correlation is removed;
Step 2, k-means clusters choose k fingerprint in whole RM and are used as initial cluster center;
Step 3, it is for other all received signal strength averages in addition to k cluster centre, then poly- with these according to them
The Euclidean distance at class center, assigns these to the cluster recently with its Euclidean distance respectively;
Step 4, has been performed after all fingerprints, obtains new cluster, using the average value of all fingerprints newly clustered as new
Cluster centre;
Step 5, continuous repeat step three and four, until k cluster centre no longer changes, terminates iteration.
2. the access point k-means clustering methods as claimed in claim 1 based on received signal strength signal ZCA albefactions, its
It is characterised by, cluster terminates, each location fingerprint is converged to cluster centre nearest therewith, each cluster is just considered as one
Individual positioning subregion;Off-line phase, each cluster and corresponding location fingerprint data constitute an independent sub- fingerprint database;
Tuning on-line stage, the received signal strength newly measured is clustered recently by being calculated with the Euclidean distance of cluster centre first
Center, is then drawn the positioning subregion of user by this cluster centre correspondence mapping function.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410212396.6A CN103987118B (en) | 2014-05-19 | 2014-05-19 | Access point k means clustering methods based on received signal strength signal ZCA albefactions |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410212396.6A CN103987118B (en) | 2014-05-19 | 2014-05-19 | Access point k means clustering methods based on received signal strength signal ZCA albefactions |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103987118A CN103987118A (en) | 2014-08-13 |
CN103987118B true CN103987118B (en) | 2017-09-05 |
Family
ID=51278911
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410212396.6A Active CN103987118B (en) | 2014-05-19 | 2014-05-19 | Access point k means clustering methods based on received signal strength signal ZCA albefactions |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103987118B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106643736B (en) * | 2017-01-06 | 2020-05-22 | 中国人民解放军信息工程大学 | Indoor positioning method and system |
CN108540926B (en) * | 2017-03-02 | 2021-01-15 | 中国移动通信有限公司研究院 | Wireless signal fingerprint construction method and device |
CN107148002A (en) * | 2017-05-27 | 2017-09-08 | 柳州天艺科技有限公司 | Primary user's localization method of RSSI based on cluster |
CN110837112A (en) * | 2018-08-16 | 2020-02-25 | 中国石油化工股份有限公司 | Data preprocessing method and system for seismic channel editing |
CN110691319B (en) * | 2019-09-03 | 2021-06-01 | 东南大学 | Method for realizing high-precision indoor positioning of heterogeneous equipment in self-adaption mode in use field |
CN110636466A (en) * | 2019-09-06 | 2019-12-31 | 联泰集群(北京)科技有限责任公司 | WiFi indoor positioning system based on channel state information under machine learning |
CN111488941B (en) * | 2020-04-15 | 2022-05-13 | 烽火通信科技股份有限公司 | Video user grouping method and device based on improved Kmeans algorithm |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20110061507A (en) * | 2009-12-01 | 2011-06-09 | 엘지전자 주식회사 | The apparatus and method for transmitting and receiving data through contention-based physical uplink data channel |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7907944B2 (en) * | 2005-07-05 | 2011-03-15 | Atc Technologies, Llc | Methods, apparatus and computer program products for joint decoding of access probes in a CDMA communications system |
-
2014
- 2014-05-19 CN CN201410212396.6A patent/CN103987118B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20110061507A (en) * | 2009-12-01 | 2011-06-09 | 엘지전자 주식회사 | The apparatus and method for transmitting and receiving data through contention-based physical uplink data channel |
Non-Patent Citations (1)
Title |
---|
Power-Efficient Access-Point Selection for Indoor Location Estimation;Yiqiang Chen, Qiang Yang;《IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING》;20060731;第4节 * |
Also Published As
Publication number | Publication date |
---|---|
CN103987118A (en) | 2014-08-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103987118B (en) | Access point k means clustering methods based on received signal strength signal ZCA albefactions | |
CN109672973B (en) | Indoor positioning fusion method based on strongest AP | |
CN106131959B (en) | A kind of dual-positioning method divided based on Wi-Fi signal space | |
US20170134899A1 (en) | Mitigating signal noise for fingerprint-based indoor localization | |
CN108696932A (en) | It is a kind of using CSI multipaths and the outdoor fingerprint positioning method of machine learning | |
CN103068035A (en) | Wireless network location method, device and system | |
CN109511085B (en) | UWB fingerprint positioning method based on MeanShift and weighted k nearest neighbor algorithm | |
CN107807346A (en) | Adaptive WKNN outdoor positionings method based on OTT Yu MR data | |
JP2000092556A (en) | Method for identifying geographic position of pattern recognition base | |
Zhou et al. | Indoor fingerprint localization based on fuzzy c-means clustering | |
CN109348416B (en) | Fingerprint indoor positioning method based on binary k-means | |
CN113596989B (en) | Indoor positioning method and system for intelligent workshop | |
CN110430578A (en) | The method for realizing cell Azimuth prediction based on mobile terminal data | |
CN106793085A (en) | Fingerprint positioning method based on normality assumption inspection | |
CN105682048B (en) | Fingerprint positioning method in Subspace Matching room based on PCA under cellular network environment | |
Zhong et al. | WiFi indoor localization based on K-means | |
CN106686720A (en) | Wireless fingerprint positioning method and system based on time dimension | |
Shao et al. | Floor identification in large-scale environments with wi-fi autonomous block models | |
CN105044659B (en) | Indoor positioning device and method based on ambient light spectrum fingerprint | |
CN103313383A (en) | Indoor positioning method based on region segmentation and curved surface fitting | |
CN107290714B (en) | Positioning method based on multi-identification fingerprint positioning | |
CN106888504A (en) | Indoor location fingerprint positioning method based on FM Yu DTMB signals | |
CN108521631B (en) | Mobile AP (access point) identification method for indoor positioning | |
CN107820202B (en) | Indoor positioning and partitioning method based on spatial characteristics | |
Dong et al. | A wifi fingerprint augmentation method for 3-d crowdsourced indoor positioning systems |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20140813 Assignee: Zhejiang Fengshou e-commerce Co.,Ltd. Assignor: ZHEJIANG NORMAL University Contract record no.: X2022980008007 Denomination of invention: K-means clustering method of access points based on zca whitening of received signal strength Granted publication date: 20170905 License type: Common License Record date: 20220623 |
|
EE01 | Entry into force of recordation of patent licensing contract |