Research On Dynamic Fingerprint Location Technology Based On RSSI

With the rapid development of Location Based Service,the precise localization of outdoor scenes has been unable to meet the demands of people towar-ds production and life,thus Indoor Localization with high accuracy has become an urgent problem to be solved.Among the various methods of Indoor Localization,Fingerprint Localization has become one of the most widely used methods because of its simplicity and high positioning accuracy.According to the uniqueness of Received Signal Strength Indication in different locations,Fingerprint Localization takes RSSI as the fingerprint that can distinguish a location uniquely.Considering that the existed Fingerprint Localization methods have poor environmental adaptability,and it is impossible to update the Fingerprint Database dynamically as well as to learn new data in an incremental way.The Dynamic Fingeiprint Localization Algorithm Frame is proposed in this paper.The main work is as fol ows:Firstly,Zigbee technology is used to build the fingerprint data collecting system,and a mobile node is introduced to collect location information as well as RSSI vector in real-time for building a Dynamic Fingerprint Database.Unlike traditional Fingerprint Localization Database in which data is collected at one time with no updating,the mobile node can update fingerprint data regularly so that the Fingerprint Database always maintains a high consistency with the current environment.Secondly,dynamic clustering is achieved which is based on Kalman Filter and Self-Organizing Incremental Neural Network algorithm.The main task of Kalman Filter is to complete the fusion between old and new fingerprint data,that is,if there are several pieces ofdata in the same position,this algorithm is used to merge them into one piece of fingerprint data.The Self-Organizing Incremental Neural Network is an online method with good incrementality and spatial proximity.It can train new data incrementally on the basis of the original training results,and gather the spatial neighboring points together through topological representation.Thus,it's particularly suitable as the clustering algorithm of Fingerprint Localization.Finally,in order to improve the positioning accuracy in online stage,Radial Basis Function algorithm is improved by a new weight-gaining method based on Kernel Density Regression.The experimental results showed that this method gained better positioning accuracy than the method of Least Square as a method of weight-gaining.A large number of experiments revealed that the Dynamic Localization Algorithm Framework proposed in this paper had good practicability.On the one hand,it can process fingerprint data dynamically,which reduces the training time of new data.On the other hand,this method maintains a better positioning accuracy compared with the original fingerprint positioning algorithm.