CN103338509A  WSN (wireless sensor network) indoor positioning method based on hidden markov models  Google Patents
WSN (wireless sensor network) indoor positioning method based on hidden markov models Download PDFInfo
 Publication number
 CN103338509A CN103338509A CN2013101216530A CN201310121653A CN103338509A CN 103338509 A CN103338509 A CN 103338509A CN 2013101216530 A CN2013101216530 A CN 2013101216530A CN 201310121653 A CN201310121653 A CN 201310121653A CN 103338509 A CN103338509 A CN 103338509A
 Authority
 CN
 China
 Prior art keywords
 node
 hidden markov
 indoor
 algorithm
 value
 Prior art date
Links
 241000134854 Aspleniaceae Species 0.000 claims description 15
 238000000034 methods Methods 0.000 claims description 12
 280000638271 Reference Point companies 0.000 claims description 8
 230000000875 corresponding Effects 0.000 claims description 6
 239000000284 extracts Substances 0.000 claims description 6
 239000011159 matrix materials Substances 0.000 claims description 3
 238000000605 extraction Methods 0.000 claims description 2
 238000005516 engineering processes Methods 0.000 description 11
 230000000694 effects Effects 0.000 description 3
 230000005540 biological transmission Effects 0.000 description 2
 241000271566 Aves Species 0.000 description 1
 238000004891 communication Methods 0.000 description 1
 230000004807 localization Effects 0.000 description 1
 238000003909 pattern recognition Methods 0.000 description 1
 239000004575 stone Substances 0.000 description 1
 239000011901 water Substances 0.000 description 1
Abstract
Description
Technical field
The present invention relates to wireless sensor network technology, artificial intelligence, fields such as pattern recognition are specifically related to a kind of WSN indoor orientation method based on hidden Markov model.
Background technology
According to " National Program for Mediumto Longterm Scientific and Technological Development (20062020))), the deployment of national science and technology development strategy planning such as " national Eleventh FiveYear Plan scientific and technical development program " and " 863 Program Eleventh FiveYear Plan development outline " file, earth observation and field of navigation technology are listed in the emphasis forward position and explore problem.Wherein, " high accuracy seamless navigation location technology " becomes the important subproblem in this field especially and paid close attention to widely.For following mobile subscriber, not only need to obtain out
Positional information under the wealthy environment, the demand to locating information under indoor environment also grows with each passing day.All like in office building,
In the indoor environments such as school, hospital, hotel, airport, railway station, warehouse, underground parking, prison, military training base, need locating information to realize the efficient management of resource in free space especially.Therefore, how to satisfy growing indoor positioning demand, become an important subject in the current earth observation research field.Along with the development of social informatization,
As the daily main activities place of people, the positional information aware services in the indoor environment has more and more stronger demand,
To greatly promote the development of China's information industry and popularize the indoor positioning Study on Technology.
In open outdoor environment, global position system GPS (Global Position System) is though can provide accurate localization information, and technology is ripe relatively, is difficult to play a role under by the indoor environment that concrete surrounded.In recent years, the high speed development of near radio power technology makes the indoor positioning rapid technological improvement.In general, indoor positioning can adopt sensor technology, transducer and the transmission network thereof of this technology by disposing in advance, and perception enters the object of specific region and finishes the location.Typical system such as infrared sensing navigation system, vibration/sound sensing positioning system, ultra broadband navigation system etc.Yet the indoor locating system that utilizes sensor technology needs the special hardware facility that increases, and equipment funds have high input, and effect is but desirable not to the utmost.The sensing range and the sensing network that are subject to transducer are disposed, and largescale positioning service can't be provided
Cover, positioning accuracy also has very big difference owing to type of sensor is different.
At present, WLAN (wireless local area network) WLAN (Wireless Local Area Network) is worldwide by widespread deployment.In view of this, Chinese scholars proposes to utilize WLAN to realize indoor positioning in succession.Like this, the user not only can enjoy conveniently information transmission of WLAN, and also can obtain positional information immediately simultaneously, thereby strengthen the function of WLAN, be to kill two birds with one stone.Problems such as and the WSN network that utilizes selforganizing carries out indoor positioning and exists: linkquality is reliable inadequately, and ambient noise is big, and the location final result shake that receives is bigger.
Summary of the invention:
The invention discloses a kind of WSN indoor orientation method based on hidden Markov model, WSN spacelocation method based on hidden Markov model, it relates to the pattern matching field, can pass through determining of HMM model, the collection of observation sequence, the training of HMM model well addresses the above problem.The present invention at first sets up complete WLAN location scene and location fingerprint storehouse; Then, the signal strength signal intensity that collects according to test point and the location fingerprint database data of prestored utilize hidden Markov model, can obtain its corresponding optimum state sequence, and then realize that the WSN interior space accurately locatees.
The present invention is achieved in that the position fixing process that it is characterized in that it:
Step 1: arrange a plurality of perception anchor nodes (Anchor Sensor Node) at indoor environment, the density requirements of deployment guarantees that the radio frequency signal that is a bit sent by one or more ASN arbitrarily that needs in the described environment to be positioned covers, and evenly plans N in described indoor environment _{RP}Individual study reference point;
Step 2: ASN gathers and gathers N respectively _{RP}The characteristic vector value of individual study reference point generates observation sequence, according to the definition of hidden Markov model and set up model Algorithm extracts the observed value sequence of node when carrying out the collection of sample, the observed value sequence is the foundation of training node locating HMM model when setting up fingerprint base; Choosing a reference point is origin Set up twodimentional rectangular coordinate system, obtain N _{RP}Individual coordinates of reference points value, according to the definition of hidden Markov model and the feature of ASN, can be divided into one of four states to ASN, respectively with node ID, node coordinate, the RSSI value, the LQI value is corresponding, here status number N=4 sets up the data fingerprint storehouse according to the signal receiver collection on each reference point from this one of four states value of each access point;
Step 3: the training node is set up hidden Markov model (HMM), earlier the data of ASN are gathered, and with its generation observed value sequence, determine one group of HMM parameter through optimizing with these observed value sequences for each node, train the hidden Markov model (HMM) of node;
Step 4: node locating.For any one mobile node (MB), we at first will carry out preliminary treatment to the characteristic value of node, do the characteristic value matching treatment with the described hidden Markov model that trains (HMM) again.Result output namely obtains the MB positional information.
The process of the training node hidden Markov model of step 3, training process is as follows:
Step 3 one: what the zone at the node place that will train was unified cuts apart, according to the top observed value sequence that is associated for each ASN extraction of telling about;
Step 3 two: set up a general HMM model Determine the status number of model, the size of the state transitions of permission and observation sequence vector;
Step 3 three: the data that training is obtained and N state carry out corresponding, calculate the initial parameter of hidden Markov model, the transition probability matrix A initialization here between the state.Set condition Can only turn back to itself or transfer to State, namely , Perhaps , namely
For the initial condition probability distribution, we set , suppose that namely HMM is from first state.And for the initialization of observing probability, we suppose: , Like this, just initially dissolve a hidden Markov model
Step 3 four: adopt at last forwardbackward algorithm or Algorithm calculates observes vector Under this model , use Cut apart replacement and evenly cut apart, carry out the initial estimation of parameter again;
Step 3 five: after initial model is determined, utilize The revaluation algorithm recomputates initial hidden Markov model, and each parameter of hidden Markov model is reappraised in this step, obtain one new , utilize then forwardbackward algorithm or Algorithm calculates the observed value sequence Under this model , in order to estimate close to the observed value sequence Model, set threshold value , when The time (at this moment Convergence), namely obtain the hidden Markov model that trains, otherwise order Repeat this step, until Convergence obtains the hidden Markov model close to the observed value sequence.
The node locating process of step 4 is supposed in the fingerprint base existing The individual node hidden Markov model that trains, its positioning algorithm based is as follows:
Step 4 one: the observation sequence vector that extracts the node that needs the location;
Step 4 two: the similar probability of hidden Markov model of each node in the observation sequence vector that calculates this node then and the node fingerprint base Similar probability Calculating be by forwardbackward algorithm or Algorithm draws;
Step 4 three: similar probability has reflected the similarity of the node hidden Markov model in node observation sequence vector to be positioned and data fingerprint storehouse, if Here it is The intermediate value maximum, so Be exactly and the immediate node hidden Markov model of node to be positioned.So far, finish the location.
Technique effect of the present invention is:This method is by the WSN of indoor existing deployment, on the basis that does not change network topology structure and systemicfunction, system increases the indoor positioning component programs, the less radiofrequency of indoor positioning mobile node (Mobile Node) is gathered the relevant characteristic parameter of distance, and utilize implicit Markov model that the characteristic vector that collects is handled, obtain Mobile Node in indoor exact position thereby overcome influences such as environmental change and artificial disturbance.
Description of drawings
Fig. 1: the structure chart that the present invention relates to hidden Markov model;
Fig. 2: the algorithm graph of a relation that the present invention relates to hidden Markov model;
Fig. 3: the hidden Markov model of WSN indoor positioning;
Fig. 4: the WSN indoor positioning procedure chart based on hidden Markov model of the present invention.
Embodiment
Further specify the WSN indoor orientation method based on hidden Markov model of the present invention below in conjunction with Fig. 1, Fig. 2, Fig. 3, Fig. 4, its concrete steps are as follows:
Training process:
Step 1: according to the required precision that node is positioned, room area to the needs location is divided grid, the size of each grid is the size of required positioning accuracy that has determined the node of location, and the positioning accuracy of node refers to the range deviation of the position that the node location oriented is actual with it.And be to require to choose according to reality to the precision that node positions, the positioning accuracy of general indoor node may be selected to be 2 meters, that is to say that the distance of the node oriented and its physical location is in 2 meters, in the present invention, the preferred square net of the shape of grid, when grid was square, the size of grid characterized with the length of side of grid, and the length of side of grid is the positioning accuracy of the node that needs the location; When grid for other shapes is, the size of grid refers to that its barycenter is to the maximum of the distance at edge.Divide the preferred evenly piecemeal of mode of grid, each piece is a grid, extracts the observed value sequence that the ASN array is associated in the network according to top telling about then.
Step 2: set up a general HMM model Determine the status number of model, the size of the state transitions of permission and observation sequence vector.According to the definition of hidden Markov model and the feature of ASN, can be divided into one of four states to ASN, respectively with node ID, node coordinate, the RSSI value, the LQI value interrelates, status number N=4 here, shown in Figure 3 is the hidden Markov model of WSN indoor positioning, wherein
ID: each node has a unique numbering, in order to distinguish the kind of node.
RSSI: full name is Received Signal Strength Indicator, is the intensity indication that receives signal, and its realization is carried out after backward channel base band receiving filter.
LQI: full name is Link Quality Indicater, IEEE 802.15.4 standard definition: indication (LQI) metering be exactly intensity and/or the quality of the packet received.Its average correlation of preceding 8 symbols that is expressed as radio communication appends in each frame of receiving together with RSSI and CRC OK/not OK.Correlation 110 expression first water frames, and be worth the minimum quality frame that 50 general expression radio detection arrive.
X, Y: be that node is at the coordinate figure of locating area.
Step 3: the data that training is obtained and N state carry out corresponding, calculate the initial parameter of hidden Markov model, the transition probability matrix A initialization here between the state.Set condition Can only turn back to itself or transfer to State, namely Perhaps , namely
For the initial condition probability distribution, we set , suppose that namely HMM is from first state.And for the initialization of observing probability, we suppose: , Like this, just initially dissolve a hidden Markov model
Step 4: adopt at last forwardbackward algorithm or Algorithm calculates observes vector Under this model With Cut apart replacement and evenly cut apart, carry out the initial estimation of parameter again.
Step 5: after initial model is determined, utilize The revaluation algorithm recomputates initial hidden Markov model.Each parameter of hidden Markov model is reappraised in this step, obtain one new Utilize then forwardbackward algorithm or Algorithm calculates the observed value sequence Under this model In order to estimate close to the observed value sequence Model, set threshold value , when The time (at this moment Convergence), namely obtain the hidden Markov model that trains, otherwise order Repeat this step, until Convergence obtains the hidden Markov model close to the observed value sequence.
Position fixing process: existing in the supposition fingerprint base The individual node hidden Markov model that trains
Step 1: the observation sequence vector that extracts the node that to identify.
Step 2: the similar probability of hidden Markov model of each node in the observation sequence vector that calculates this node then and the node fingerprint base Similar probability Calculating be by forwardbackward algorithm or Algorithm draws.
Step 3: similar probability has reflected the similarity of the node hidden Markov model in node observation sequence vector to be positioned and data fingerprint storehouse, if Here it is The intermediate value maximum, so Be exactly and the immediate node hidden Markov model of node to be positioned.So far, finish the location.
Claims (3)
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

CN2013101216530A CN103338509A (en)  20130410  20130410  WSN (wireless sensor network) indoor positioning method based on hidden markov models 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

CN2013101216530A CN103338509A (en)  20130410  20130410  WSN (wireless sensor network) indoor positioning method based on hidden markov models 
Publications (1)
Publication Number  Publication Date 

CN103338509A true CN103338509A (en)  20131002 
Family
ID=49246588
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

CN2013101216530A CN103338509A (en)  20130410  20130410  WSN (wireless sensor network) indoor positioning method based on hidden markov models 
Country Status (1)
Country  Link 

CN (1)  CN103338509A (en) 
Cited By (11)
Publication number  Priority date  Publication date  Assignee  Title 

CN103796304A (en) *  20140115  20140514  内蒙古科技大学  Coal mine underground positioning method based on virtual training set and Markov chain 
CN104144495A (en) *  20140704  20141112  中国科学院光电研究院  Fingerprint positioning method based on direction sensor and WLAN network 
CN104270796A (en) *  20141023  20150107  湘潭大学  Data collection method for selecting Sink routes based on markov model 
CN104469942A (en) *  20141224  20150325  福建师范大学  Indoor positioning method based on hidden Markov model 
CN106028290A (en) *  20160506  20161012  浙江工业大学  WSN multidimensional vector fingerprint positioning method based on Kriging 
CN106371064A (en) *  20160908  20170201  成都希盟泰克科技发展有限公司  Positioning method based on layered hidden Markov model (HMM) 
CN106597363A (en) *  20161027  20170426  中国传媒大学  Pedestrian location method in indoor WLAN environment 
CN107003382A (en) *  20141217  20170801  索尼公司  Message processing device, information processing method and program 
CN107135244A (en) *  20160229  20170905  阿里巴巴集团控股有限公司  Locationbased service implementation method and device 
CN109982242A (en) *  20190329  20190705  深圳市九洲电器有限公司  A kind of indoor orientation method, device, base station and system 
US10849205B2 (en)  20151014  20201124  Current Lighting Solutions, Llc  Luminaire having a beacon and a directional antenna 
Citations (2)
Publication number  Priority date  Publication date  Assignee  Title 

CN1666112A (en) *  20020531  20050907  埃卡豪股份有限公司  Sequencebased positioning technique 
CN101526604A (en) *  20080307  20090909  财团法人工业技术研究院  device and method for converting signal intensity of wireless positioning system 

2013
 20130410 CN CN2013101216530A patent/CN103338509A/en not_active Application Discontinuation
Patent Citations (2)
Publication number  Priority date  Publication date  Assignee  Title 

CN1666112A (en) *  20020531  20050907  埃卡豪股份有限公司  Sequencebased positioning technique 
CN101526604A (en) *  20080307  20090909  财团法人工业技术研究院  device and method for converting signal intensity of wireless positioning system 
NonPatent Citations (1)
Title 

OBAID UR REHMAN, ET AL: "Performance study of localization techniques in wireless body area sensor networks", 《ARXIV》 * 
Cited By (17)
Publication number  Priority date  Publication date  Assignee  Title 

CN103796304B (en) *  20140115  20170630  内蒙古科技大学  One kind is based on virtual training collection and markovian underground coal mine localization method 
CN103796304A (en) *  20140115  20140514  内蒙古科技大学  Coal mine underground positioning method based on virtual training set and Markov chain 
CN104144495A (en) *  20140704  20141112  中国科学院光电研究院  Fingerprint positioning method based on direction sensor and WLAN network 
CN104144495B (en) *  20140704  20160511  中国科学院光电研究院  A kind of fingerprint positioning method based on direction sensor and wlan network 
CN104270796B (en) *  20141023  20171212  湘潭大学  Method of data capture based on Markov model selection Sink paths 
CN104270796A (en) *  20141023  20150107  湘潭大学  Data collection method for selecting Sink routes based on markov model 
CN107003382A (en) *  20141217  20170801  索尼公司  Message processing device, information processing method and program 
CN104469942A (en) *  20141224  20150325  福建师范大学  Indoor positioning method based on hidden Markov model 
CN104469942B (en) *  20141224  20180227  福建师范大学  A kind of indoor orientation method based on HMM 
US10849205B2 (en)  20151014  20201124  Current Lighting Solutions, Llc  Luminaire having a beacon and a directional antenna 
CN107135244A (en) *  20160229  20170905  阿里巴巴集团控股有限公司  Locationbased service implementation method and device 
US10904707B2 (en)  20160229  20210126  Advanced New Technologies Co., Ltd.  Locationbased service implementing method and apparatus 
CN106028290A (en) *  20160506  20161012  浙江工业大学  WSN multidimensional vector fingerprint positioning method based on Kriging 
CN106371064B (en) *  20160908  20181120  成都希盟泰克科技发展有限公司  A kind of localization method based on layering Hidden Markov Model 
CN106371064A (en) *  20160908  20170201  成都希盟泰克科技发展有限公司  Positioning method based on layered hidden Markov model (HMM) 
CN106597363A (en) *  20161027  20170426  中国传媒大学  Pedestrian location method in indoor WLAN environment 
CN109982242A (en) *  20190329  20190705  深圳市九洲电器有限公司  A kind of indoor orientation method, device, base station and system 
Similar Documents
Publication  Publication Date  Title 

Xia et al.  Indoor fingerprint positioning based on WiFi: An overview  
TorresSospedra et al.  UJIIndoorLoc: A new multibuilding and multifloor database for WLAN fingerprintbased indoor localization problems  
CN105588566B (en)  A kind of indoor locating system merged based on bluetooth with MEMS and method  
CN103561462B (en)  Indoor positioning system and method totally based on smart mobile terminal platform  
CN104796866B (en)  Indoor orientation method and device  
Shu et al.  Gradientbased fingerprinting for indoor localization and tracking  
CN103842840B (en)  The method of mobile device and operation mobile device  
CN103476115B (en)  A kind of WiFi fingerprint positioning method based on AP collection similarity  
CN102098780B (en)  Positioning method and device  
CN103363988B (en)  A kind of method utilizing intelligent mobile phone sensor to realize the navigation of earth magnetism indoor positioning  
CN105044662B (en)  A kind of fingerprint cluster multipoint joint indoor orientation method based on WIFI signal intensity  
CN103728645B (en)  A kind of indoor and outdoor stationkeeping system and localization method thereof  
CN105474031B (en)  The 3D sectorization path loss models that 3D for mobile terminal is positioned  
CN103338516B (en)  A kind of wireless sensor network two step localization method based on total least square  
CN105137390B (en)  A kind of indoor orientation method based on adjustable transmission power AP  
CN103200678B (en)  The Android device WiFi indoor orientation method of positionbased algorithm for recognizing fingerprint  
CN104602342B (en)  A kind of high efficiency indoor localization method based on iBeacon equipment  
CN104215238A (en)  Indoor positioning method of intelligent mobile phone  
Jang et al.  Indoor positioning technologies without offline fingerprinting map: A survey  
Peng et al.  An iterative weighted KNN (IWKNN) based indoor localization method in bluetooth low energy (BLE) environment  
Zhang et al.  A comprehensive study of bluetooth fingerprintingbased algorithms for localization  
CN101868026B (en)  Multihop positioning method for lightweight wireless sensor networks  
CN103533647B (en)  A kind of radio frequency map adaptive location method based on subclustering mechanism and robustness regression  
CN106793086B (en)  Indoor positioning method  
CN102170697B (en)  Indoor positioning method and device 
Legal Events
Date  Code  Title  Description 

C06  Publication  
PB01  Publication  
C10  Entry into substantive examination  
SE01  Entry into force of request for substantive examination  
C02  Deemed withdrawal of patent application after publication (patent law 2001)  
WD01  Invention patent application deemed withdrawn after publication 
Application publication date: 20131002 