Indoor Positioning Method Based On The RSSI Statistical Probability Distribution

With the popularization and development of wireless network technology and computer applications, people can use the terminal device to connect wireless hotspots easily. Get the result by using the difference of RSSI in different locations because of large number of wireless local area network hotspots in public places. WLAN indoor positioning method is widely used in indoor positioning system by using the wireless access points and could get a better positioning accuracy. The current WLAN indoor positioning method is mainly divided into two categories:Propagation model method and Fingerprint method. The fingerprint method is widely used for its stability and practicability in engineering. Fingerprint method is mainly divided into two stages:the offline stage and the online stage. Major works of offline stage is the collection of RSSI, data-processing and modeling. However, data-processing of positioning point and get the location by matching algorithm is the major work of online stage.Current methods of WLAN indoor positioning is mainly to improve the positioning algorithm or use new filtering algorithm to get a higher positioning accuracy due to the presence of the signal variability affects the accuracy of positioning. This thesis proposed several improved indoor positioning method based on RSSI statistical probability distribution and introduced Support Vector Machine classification, improved the positioning accuracy and efficiency.Firstly, fingerprint method is used in this thesis. Collect the data of RSSI in fingerprint points, and establish the multivariate fitting Gaussian distribution model and SVM classification model in offline phase. Secondly, collect the real-time data of positioning point and get the multivariate fitting Gaussian distribution model of positioning point in online phase. Classified by SVM classification model and matched by fitting Gaussian distribution model. Finally, obtain the location with WKNN algorithm by assigning different weight values. Experimental results show that the proposed methods not only improve the positioning accuracy but also improve the positioning efficiency, enhance the robustness and practicability.