Research On Lightweight Indoor Fingerprinting Localization Methods Based On CSI

In recent year,with the development of the mobile Internet,the demand for location-based services has been increasing continuously.At present,indoor positioning technology has yet to be developed.The indoor positioning based on Wi Fi has the advantages of being widely deployed,which has attracted widespread attention from researchers.At present,researchers can use commercial network cards to obtain CSI in Wi Fi signals.Using CSI as positioning reference information,has positioning accuracy advantages.At present,indoor positioning using CSI as positioning reference information has the problems of high labor cost and high hardware cost,which is not suitable for widespread deployment.In summary,in order to reduce the cost of building an indoor positioning system and achieve a certain positioning accuracy under the condition of reducing costs,this thesis proposes a lightweight indoor positioning method based on CSI.First,the thesis proposes an offline data processing method for indoor positioning.By continuously collecting fingerprint data of different position points,the fingerprint point separation algorithm is used to separate the data of different fingerprint points.This method can reduce the time to collect fingerprint data offline.After separating the fingerprint points,in order to select the access point with better positioning effect,a method for evaluating the consistency of the fingerprint data of the base station is designed.This method calculates the average degree that each subcarrier deviates from its center amplitude in the CSI data,and determines the final retention of the fingerprint data by the evaluation value.This method reduces unnecessary deployment of base stations and can reduce the hardware cost of building an indoor positioning environment.At the same time,the amount of data stored in the fingerprint database is reduced,and the data stored in the fingerprint database is relatively stable.Then,the thesis designs an improved WKNN location fingerprint matching algorithm.Based on WKNN,this method corrects the distance ranking to reduce the impact of long distance similar points on positioning accuracy.Considering the situation where there is more interference indoors or the multipath effect is rich,the thesis proposes a method of iterative feature extraction of the collected CSI data using a random forest model.In order to obtain several features that are more effective in distinguishing locations when there is interference indoors.The extracted data is used for online matching to improve the robustness of the positioning system.Finally,the proposed lightweight positioning method is experimentally verified.Using a small indoor positioning scene as the experimental environment to restore the indoor positioning scenes common in daily life.In the experimentally scene,compare the positioning accuracy of the indoor positioning method proposed in the thesis with the existing methods,and compared with the accuracy of positioning directly using a random forest classification system.The experimental results show that the positioning algorithm and data processing method proposed in this thesis can achieve better positioning accuracy on the premise of satisfying the lightweight.