The Research On Radio Frequency Based Device-free Indoor Localization

With the development of wireless sensor networks and intelligent devices,LBS is playing an increasingly important role in real life.These services are widely used in many fields such as travel navigation,security and rescue,and health care.How to accurately obtain the target's location information has become a hot topic of domestic and foreign scholars,a variety of localization technology have emerged.Compared to the outdoor localization,indoor localization is facing many challenges: complex indoor environment,higher accuracy requirement etc.Therefore,the exploration and research on more stable and accurate indoor localization technology has certain practical value.In this dissertation,a radio frequency based device-free indoor localization model is constructed.The principle of the localization model is to use the different effects on the received signal strength caused by different locations to finish the localization in the indoor environment.This device-free localization technology does not need people to carry any equipment,especially for the elderly living alone in the indoor localization and care.Then,the localization process is optimized by combining the position continuity characteristics of the moving target.Based on the model,in order to improve the accuracy of localization,this dissertation proposes a class-based feature selection method in the indoor localization,which transforms the multi-classification into many binary classifications.This feature selection method can select feature subsets owning discriminant ability for each class and map these feature subsets to the original data set,and the data set after dimension reduction is obtained.In the classification phase,using the SVM method to train classifier models,our method can predict the location of testing sample combined with the probability estimates and finish the localization.The experimental results show the device-free localization model is robust under different experimental settings,and the localization accuracy is higher when the SVM is used as the localization classification algorithm.In the further experiment,the feature selection method in this dissertation can effectively eliminate the irrelevant and redundant features,reduce the feature dimension of the original set and improve the accuracy of localization,so that the average localization accuracy reaches 96.11%.