Dynamic Sitting Posture Recognition System Using Passive RFID Tags in Internet of Things

Abstract

The rapidly developing Internet of things technology is gradually being utilized to monitor people's unhealthy behaviors. Sedentary behavior and wrong sitting posture are common health issues that can detrimentally impact the physical and psychological health of teenagers. An effective way to promptly rectify improper sitting postures among teenagers is to use equipment to monitor and recognize the alterations of sitting posture. The majority of conventional sitting posture recognition methods rely on cameras or sensors to recognize sitting posture. The employment of cameras will violate user's privacy, and the utilization of sensors will increase the cost of implementation. A dynamic sitting posture recognition system based on commodity off-the-shelf (COTS) radio frequency identification (RFID) devices is proposed. This system can recognize six common erroneous sitting postures by simply sticking five passive RFID tags on the user's back. We collect phase and received signal strength indicator (RSSI) data of passive RFID tags, then transform them into Doppler shift and RSSI difference data respectively, and finally input them into the established deep residual neural network for the classification of sitting postures. The experiment results show that our system achieves an average recognition accuracy of 99.17% with six sitting postures and is highly robust to different users and different usage environments.