Research On Indoor WiFi Localization Algorithm With CSI Fingerprint

Location based service has become an essential part of daily life.Indoor localization system based on Wi Fi has become research hotspot in recent years due to its low cost and simple installation.The traditional WiFi indoor localization method usually adopts the received signal strength(RSS)as the parameter.,but RSS can be easily affected by the environment,which will lead to low position accuracy.With the application of OFDM and MIMO technology in WiFi standard,it is possible to obtain the channel state information(CSI)of the physical layer,and CSI is a parameter which reflects the finer granularity of the channel state than RSS.Therefore,the current indoor location research has turned to use CSI instead of RSS as fingerprint information to improve accuracy.In this thesis,we focus on the WiFi fingerprint indoor localization algorithm based on CSI,and fully mining the rich information of CSI.The main research content of the thesis are summarized as follows:(1)The basic theory of WiFi indoor localization technology is introduced,and the advantage of using CSI instead of RSS as fingerprint information is pointed out.(2)The CSI fingerprint library is established by using the principle component of CSI sample set in offline phrase.Firstly,a density-based Spatial clustering of applications with noise algorithm is applied to eliminate the outlier noise sample.Then,in order to make use of CSI phase information,a real time phase information is extracted based on linear transformation.Finally,in order to reduce the complexity of the algorithm and improve the fingerprint validity,principal component analysis is proposed to extract the main eigenvalues of the CSI sample set matrix as the fingerprint.(3)Combined with practical problems,k nearest algorithm and support vector regression method are applied in online phase.Thereby,the location coordinates are predicted according to its fingerprint information.(4)Experiments were carried out in two indoor environments of the hall and restaurant.The effectiveness and localization accuracy of the indoor location method proposed in this thesis were verified by analysis.Then compare with the indoor positioning system proposed by CSI-MIMO.The experimental results show that the average error of the PCA-kNN indoor localization algorithm based on the main eigenvalues of the CSI sample set is 1.36 m,and the accuracy of the kNN algorithm is 11% higher than that of the CSI-MIMO system.The average error of the support vector machine regression method is 1.44 m,and the accuracy of the CSI-MIMO system is improved by 6%.In a restaurant with a more complex indoor environment,The average positioning error of PCA-kNN algorithm and SVR algorithm is 1.73 m and 1.58 m respectively,and the positioning accuracy of the CSI-MIMO system is 12% and 20% respectively,which are increased by 12% and 20% respectively compared with the CSI-MIMO system.