Towards Accurate Indoor Localization Using Channel State Information

With the popularity of wireless communication networks and intelligent devices,location based services have been getting into people's lives deeply,which promote the researchers and the industries give more energy and enthusiasm into the study of localization.Indoor positioning,due to its inherent limitations,still lack of effective positioning technology.In this paper,based on the full study of a variety of positioning technology,I propose several indoor positioning methods based on channel state information.First of all,this paper focuses on WiFi based indoor positioning technology,mostly on received signal strength and channel state information positioning technology.Based on the widely reading of papers,this paper points out the lack of the RSSI positioning technology and the advantages of channel state information.This paper studies the indoor positioning algorithm based on channel state information and analyzes the problems of existing algorithms.Through experiments and data analysis,this paper expounds the stability of channel state information with time factor,the variability of the direction of receiving equipment,the fluctuation of personnel movement,the sensitivity of position change and demonstrates the position recognition effectiveness.At the same time,this paper points out the key point while using channel state information to position and the impact of different factors.Next,in order to remove the correlation between different subcarrier signals,a normalization method is proposed.In order to fully extract the difference of different training samples and reduce the influence of instantaneous fluctuation signals,this paper proposes to reduce the dimension of original data and extract useful feature by using weighted principal component analysis algorithm,Combining weighted principal component analysis algorithm with bayes algorithm to locate the unknown data.In the existing literature,the na?ve bayes algorithm is used to assume that all the data obey a certain distribution.In order to avoid doing such hypothesis,a statistic bayes algorithm is proposed.Next,in this paper,I use the linear discriminant analysis algorithm based on Fisher criteria function to extract feature which is supposed to enlarger the similarity between classes and decrease the difference in a certain class,and combine it with bayes algorithm to achieve higher positioning accuracy.In order to reduce the influence of environmental factors such as human movement,the concept of "Location Image" is introduced,thus using the two-direction two-dimension principal component analysis algorithm to remove the correlation of the "Location Image".The bayes algorithm and K-nearest neighbor algorithm are used to locate the position.In this way,not only the computational complexity is reduced but also the accuracy is improved.Finally,the experiment with the Intel 5300 wireless network card and NETGEAR were carried out in two experimental scenarios,and the validity of the proposed algorithm was verified.