R&P: An Low-Cost Device-Free Activity Recognition for E-Health

Activity recognition is important for taking care of patients and old men especially in e-Health. The activity recognition system without carrying any wearable devices is widely used in our daily life. Current methods employing uneconomical equipment or even dedicated devices lead to cost-inefficiency for large-scale deployments. This paper introduces R&P, a device-free activity recognition system only using cheap radio frequency identification devices (RFID) tags. Based on the analysis of RFID signals, we extract received signal strength fingerprints and phase fingerprints for each activity and synthesize these two kinds of fingerprints to accurately recognize activities. Moreover, we also modify the dynamic time warping (DTW) algorithm and propose T-DTW method to improve the recognition efficiency. We use commercial passive RFID hardware and verify R&Pin three different environments with different targets and six activities. The results demonstrate that our solution can recognize activities with an average accuracy of 87.9%.