T-testing Of Location Fingerprint For WLAN Indoor Localization

The wide application of Wireless Local Area Network(WLAN)and mobile devices drives the increasing demand for the user Location Based Service(LBS).Therefore,indoor localization based on Received Signal Strength(RSS)has become one of the hotspots of scientific research.Based on the previous WLAN indoor localization algorithms,T-testing for location fingerprint is proposed.Specifically,the thesis about sample capacity minimization is proposed.Morever,a statistical hypothesis testing based fingerprint matching localization method is designed and implemented to achieve the localization estimation for target terminals.In all,three contributions of this thesis are listed as follows:First of all,the research on WLAN localization fingerprint database optimization is completed.In offline phase,based on the observation of RSS fluctuation,the Operating Characteristics(OC)function is utilized to obtain the allowable minimal RSS sample capacity for fingerprint database construction.At the same time,by the concept of information gain,the optimal Access Point(AP)combination which can provide the best localization performance is selected to establish the database.This method effectively resolves the blindness and unreliability of localtion fingerprint data collection,and makes the difference of the RSS data between different physical locations in the target environment more obvious.In addition,the proposed approach reduces the data processing overhead as well as decreases the time cost.Then,through the homogeneity of variance test,the RPs with same variance for each Test Point(TP)is inferred.The variance reflects the RSS samples' deviation level from their mean value,as well as the fluctuation degree within the RSS samples.Based on this,a hypothesis testing model are constructed to calculate the similarity between the online RSS sample sequence and the offline RSS sample sequence collected at each AP,and the corresponding statistics as well as reject region are also determined.After that,the homogeneity of variance with respect to the RSS sample sequence from each AP is analyzed,and the RPs with same variance are regarded as a coarse localization result.Finally,based on comprehensive experiments,it has been verified that this method can effectively match the TPs to its belonging sub-regions.In the end,the localization of target terminal is realized by the outlier detection and the T-test matching.The Density-based Spatial Clustering of Applications with Noise(DBSCAN)algorithm is proposed to eliminate the outliers of physical locations which are not in the regions with same-density.Then,according to various digital features including the mean value and size of RSS samples,the similarity between RSS samples is analyzed.Finally,the statistical hypothesis T-test is performed to complete the precise localization of target terminal.The experimental simulation shows that the stability of the localization performance can be well guaranteed with the reduction of time cost.