Research On Indoor Positioning Technology Based On WiFi Channel State Information

With the development of wireless communication technology and the popularization of Wi Fi,wireless local area networks have been widely used in indoor environments,which promotes indoor localization technology based on Wi Fi has also developed.Wi Fi-based indoor localization technology is mainly developed based on Wi Fi channel state information(CSI),compared with the MAC layer information provided by RSS,it can provide rich physical layer information,which contains rich features,such as the CSI amplitude and phase of each subcarrier,etc.,which laid the foundation for CSI indoor localization.At present,indoor localization technology is mainly divided into two types in the localization phase: geometric-based localization and feature-based localization.The latter method is adopted in this paper.The fingerprint database is constructed by selecting appropriate feature parameters to complete the location estimation of the nodes to be located.The main research contents of this paper are as follows:First,according to the LOS and NLOS factors that affect localization accuracy,the CSI amplitude matrix is converted into a distance matrix as a characteristic parameter,the LOWESS regression model of Rice-K factor and antenna phase difference variance factor is established to calculate the distance corresponding to each subcarrier in different environments,which can be used to obtain CSI distance matrix.It overcomes the problem of the amplitude matrix failure caused by NLOS and lays the foundation for the positioning phase.Then,in order to obtain stable fingerprint database characteristic parameters,support vector machine(SVM)based burst disturbance detection technology was used before the fingerprint database was built,which mainly used the presence of sudden disturbances,such as the impact of human movement on the statistical characteristics of the amplitude of CSI as the basis for judgment.Through research and analysis,it is found that the variance of the CSI amplitude of different subcarriers under burst disturbance is greater than that without burst disturbance,so the amplitude variance of 30 subcarriers is used as the feature vector for SVM discrimination to calculate the detected hyperplane and intercept;After that,in order to reduce the number of matches during the online localization phase,a hierarchical clustering method was used to cluster the fingerprint database according to the characteristics of the distance matrix.The fingerprint database was divided into several clusters,the center of each cluster is first compared during the localization matching and the cluster with the highest similarity center is selected as the matching area,then the Gaussian Weighted KNN matching algorithm is used to complete the localization estimation of the node to be located.Finally,two experimental scenarios were selected to evaluate the indoor localization system.One PC equipped with Intel 5300 wireless network card was used as the receiver,and the Tenda router was used as the transmitter.The experimental data and algorithms are simulated by using MATLAB and Python analysis tools.And simulation results show that the distance matrix achieved good localization accuracy in both experimental scenarios.