Research Of Indoor Location Method Based On Channel State Information

With the rapid development of information industry in modern society,locationbased information service has been paid more attention.At the same time,diversified application scenarios put forward higher requirements for accuracy and stability of location services.As a common basic network facility,Wi-Fi devices have been widely deployed in daily life scenarios,so the indoor positioning method based on Wi-Fi has a great advantage in promotion.The traditional indoor positioning technology based on Wi-Fi uses RSSI as the reference quantity of position resolution,but RSSI has poor anti-interference performance and is seriously affected by multipath effect,so it is difficult to meet the requirements of high precision and high reliability.The channel state information(CSI)of IEEE 802.11 n protocol Wi-Fi physical layer records the power attenuation and phase offset of each sub-carrier in the signal propagation process in detail.Compared with RSSI,it describes the multipath propagation process in a finer granulation,which has the conditions to further improve the accuracy of the positioning system.However,full CSI is a high-dimensional nonlinear data,which is difficult to be directly used for location analysis.In this dissertation,the algorithm and application of CSI fingerprint indoor positioning technology were studied.In order to make full use of the rich signal transmission process information of CSI and ensure the high efficiency of the positioning system,an effective positioning method was proposed through feature extraction and location subspace division,and a real-time CSI indoor positioning platform was finally built.The main work is as follows:(1)In order to solve the problems of big data redundancy and complex analysis of CSI fingerprint in indoor location applications with spatial differences,representation learning was introduced into the feature extraction task of fingerprint location.A CSI indoor location method was proposed based on Stacked Sparse Auto-encoder(SSAE)Network combined with kernel Support Vector Machine(SVM).The amplitude and phase offset of the single AP wireless subcarrier in the process of spatial propagation were collected in full quantity,and the location fingerprint was generated by the CSI amplitude and phase information simultaneously.After the amplitude and phase information of CSI is preprocessed and input to SSAE,the output vector which tends to be sparse in the encoder part is taken as the low-dimensional representation of the initial fingerprint to establish the fingerprint database.Fingerprint database is used as model training set,and the mapping relationship between fingerprint and actual physical location is established through kernel SVM.Experimental results show that our method can achieve high precision position resolution in complex environment.(2)In order to further improve the efficiency of the positioning system,a CSI indoor positioning method based on fuzzy delimit molecular space was proposed on the basis of sparse characteristic fingerprint database.The whole location space is divided into several cross-subspaces by the Fuzzy C-means clustering algorithm optimized by particle swarm optimization(PSO).The initial cluster center of the fuzzy clustering algorithm is obtained through PSO optimization,and then further iterative tuning is carried out to ensure the reliability of clustering results.Real-time fingerprint matching process is completed inside the subspaces,which greatly reduces the system resources consumed by the location task.In the subspace,the weighted k-nearest neighbor algorithm(WKNN)is improved to complete the fingerprint matching calculation,and the fuzzy membership is introduced to further improve the accuracy of fingerprint matching.The experimental results confirm that our method has better positioning accuracy and real-time performance in the complex laboratory environment and corridor environment with fewer obstacles,which confirms the effectiveness of the proposed method.(3)This topic designed a CSI indoor positioning application platform and displayed the output results of the positioning algorithm under the scene.The positioning platform is built based on B/S structure and developed by Spring Boot framework.The overall design of user management,single point positioning,historical position function modules and database tables is completed,and then the functions of user login,real-time single point positioning display and historical position track query are realized.