Research Of Indoor Localization And Tracing Technology Based On CSI

With the rapid development of wireless networks,people's needs for all aspects of life and work are getting higher and higher.Indoor localization services are increasingly needed by people for their wide practicality and functionality.Since Wi-Fi devices are common in public places,offices,or private residence,Wi-Fi-based indoor localization is the focus of today's research.Common research in the Wi-Fi environment is based on Received Signal Strength Indication(RSSI),but the effect of localization is not ideal.The Channel State Information(CSI)that can be obtained from the wireless network card is a more elaborate feature,and it is more sensitive to the perception of the environment.In this paper,the amplitude of the Channel State Information is used as the feature value to design the localization and tracking algorithm.Existing localization systems usually use Support Vector Machine(SVM),K-Nearest Neighbor and other classification methods to complete localization.These methods are characterized by their complexity and complex fingerprint database.Some systems use bayes classifier,but they ignore the correlation between eigenvalues,which leads to low accuracy.In order to solve these problems,this paper proposes an indoor localization method combining Principal Component Analysis(PCA)and Naive Bayes Classifier(NBC),namely,naive bayes localization algorithm with independent attributes,which has the advantages of simple fingerprint database,high positioning accuracy and fast speed.It is divided into two stages.In the off-line training stage,principal component analysis is first used to reduce the dimensionality of the CSI data to make it become independent.We modeled the eigenvalues of each location as normal distribution,so the mean and variance of the new eigenvalues were extracted and stored in the fingerprint database.In the online testing stage,the attribute-independent weighted naive bayes classification method is used.The test data is also processed by principal component analysis,and the variance contribution rate of each principal component is recorded as the weight of the classification,which will be brought into the classifier later.In terms of tracking,Kalman filtering is often used,but the tracking effect is not good for uncertain target motion state.Therefore,this paper proposed a tracking algorithm based on simplified Sage-Husa adaptive filtering,and combined with the localization results to complete target trajectory tracking.This algorithm takes the localization error calculated in the localization stage as fixed observation noise,and uses time-varying noise statistical estimator to constantly modify the system noise,so that the system equation is more consistent with the actual motion state.Finally,the accuracy and performance of the localization algorithm and the tracking algorithm are verified by experiments.