Research On Indoor Context-aware Techniques Based On RFID

With the development of "Internet of Things",near-field communication technology supported by wireless sensor network technology and RFID technology has become increasingly mature.Context-aware Techniques Based on RFID becomes increasingly popular due to its high precision,low cost and Ubiquity.Physical layer information such as RSSI and phase can be obtained from COTS RFID devices,but the raw data is unstable and contains many noise signals.Therefore,new theory and technologies are needed for environmental perception.In this thesis,we extract more stable RF features to achieve more robust perception and take three typical scenarios as examples.In the first scene,we simplify daily activities of the elderly into fast walking,slow walking,sitting down,and fall.We extract features by the sliding window and phase fitting method and conduct activity recognition by DTW and SVM.Finally,a simple health monitoring system is realized for the elderly.In the second scenario,we track RFID tags by the received reflected signal.In view of the poor stability and weak anti-interference ability of traditional RSSI,this paper introduces LWLR algorithm and introduces phase information to guide person's moving direction.Finally,we achieve a blind anti-collision warning system with more robust performance.The third scenario is mainly aimed at reducing the occurrence of punching phenomenon.The frequency distribution histogram is used to extract the features and kmeans algorithm is used to reduce the false positive rate of the recognition of similar people based on the strong anti-interference ability and non-intrusive characteristics of RFID signals and thus a more robust RFID-based attendance system is finally achieved.At last,a large number of experiments are conducted and verify our three systems.Experiment results show that the RFID-based system has high performance and availability.