Research On WLAN Indoor Positioning Technology Based On Location Fingerprinting

The Location Based Service(LBS)is increasingly demanded due to the development of wireless communication technology and the universality of smart mobile device.Therefore,the research on indoor positioning system is paid broad attention to,and has huge business potential and market value.However,Global Positioning System(GPS)cannot work effectively indoors because of so many factors,such as the complexity of the indoor environment and non-line of sight propagation of signal.Then,based on the characteristic of wide-used Wireless Local Area Network(WLAN)in buildings,which is that the Received Signal Strength(RSS)changes for different positions,the research on indoor positioning technology based on location fingerprint is hot in the LBS field.This thesis mainly analyzes WLAN indoor positioning technology based on location fingerprint,and is divided into three parts.Firstly,the selection of Access Point(AP)is discussed.The RSS signal is collected in real indoor environment and mainly analyzed its propagation feature of time and spatial dimension.Because of the issues like obstruction and multipath effect,the RSS signal detected at Reference Point(RP)is extremely unstable.Thus,saving all RSS signal in fingerprint database influences the accuracy of positioning system.Accordingly,the improved AP selection algorithm is introduced based on the mean value and stability of signals to select the optimal subset of APs as fingerprint data.And it is verified by the tests that the accuracy of indoor positioning is improved because of the AP selection algorithm introduced in this thesis.Secondly,the feature extraction of RSS is analyzed.Feature extraction can be used for eliminating redundancy and noise,reducing the computational complexity and enhancing the positioning accuracy,since the storage and computation of mobile devices are limited but the information in fingerprint database is quite a lot.Traditional ways of feature extraction only analyze the linear relation among different data.Therefore,this thesis applies Kernel Principle Components Analysis(KPCA)for mapping the lowdimensional data to high-dimensional for nonlinear feature extraction,which can further increase the positioning performance compared to the traditional algorithm.Finally,the indoor positioning model based on clustering is studied.The RSS changes hugely for the spatial distribution in buildings of large area.It is not conducive to improving the performance of positioning system if the computational quantity increases by analyzing the signal of the whole area.Hence,based on clustering,this thesis pays much attention on analysis and improvement of indoor positioning model.The Affinity Propagation Clustering(APC)algorithm is exploited to divide positioning environment into various areas,and similarity matching algorithm is raised to make sure the target area.The exact location of the target is determined in each small area.The clustering algorithm enhances the positioning accuracy with the computational complexity reduction.