Multi-Person Pose Recognition Using a Zigbee Sensor Network

Abstract

This paper presents a novel design of a robotic multi-person posture recognition system. The proposed system consists of a human posture detection module, a wireless sensor network and a multi-person activity monitoring software running on a mobile robot. The human posture detection is based on a triaxial accelerometer and the developed posture recognition algorithms. Further, the posture detection module has a Zigbee chip and an 8-bit microcontroller on board. Thus, several posture detection modules and a Zigbee node connected to the robot control computer form a Zigbee wireless network. In the Zigbee sensor network, each posture detection module can communicate with the robot onboard computer. The multi-person activity monitoring software can monitor and record the postures of multiple users on-line in real time. A posture classification algorithm is proposed by combining time-domain analysis and wavelet transform analysis. The complete algorithm has been realized in the microcontroller of the human posture detection module. In the current design, the system can classify seven human postures: falling, standing, sitting, lying, walking, downstairs. After testing the system by five users, we have demonstrated an accurate rate of 88%.