Research On Indoor Localization Technology Based On WIFI

The Location Based Service(LBS)has attracted more and more attention in recent years with the improving demand of positioning accuracy caused by the fast development of artificial intelligent technology.Although the outdoor location system is performing well with the help of satellite signal,survey shows that there are no universal indoor localization systems with high accuracy.The indoor localization technology is becoming the hot research point because of its low cost and low consumption of energy.Based on the RSS fingerprint technology of WIFI signals in indoor localization systems,this paper promotes the RSS fingerprint localization algorithm,and designs a verification system on Android platform.The detailed studies are as follows:Firstly,we discover that the traditional near neighbors' selection algorithm used fingerprint information limited and unable to remove singular points efficiently,which could affect the location accuracy seriously.To solve this problem,we added the influence of AP matching degree.We optimize the traditional algorithms by proposing the Similarity based Near Neighbors Selection algorithm.Referring to the idea of solving similarity of nodes in complex networks,this strategy defines a parameter named "similarity" based on the principle of Jaccard similarity coefficient to eliminate the effects cause by singular points.Secondly,the extracting process of traditional RSS vector would easily lead into large volatility signal which delivers inaccurate RSS vector collected by TP and RP.Thus,this paper proposes the Cosine Similarity-based AP Selection Algorithm to remove larger volatility vectors through computing cosine values between vectors.Finally,the indoor localization system based on Android platform is realized,which is used to evaluate the performance of proposed algorithm in a real indoor Wi-Fi environment.The experimental results show that the similarity degree based on Near Neighbors Selection algorithm has 12% higher positioning accuracy comparing with traditional near neighbors selection algorithms,and the Cosine Similarity-based AP Selection Algorithm improves the accuracy of RSS vectors,which makes the positioning accuracy 7.8% higher than that with the maximum mean method.