A Real-time Device-free Head Motion Recognition Framework for Family Care Robots

In this paper, we propose a practical framework to realize accurate, real-time head motion recognition without wearing a device on user's head, which is essential for interaction between human and family care robot system. Most of the previous work has focused on the estimation of head poses rather than the identification of specific head motions, which are more useful in human-robot interaction. Based on the results of the head pose estimation, we define eight head motions which contain most of the head movements we use in daily life and design a finite state machine(FSM) to identify these motions. Our framework shows excellent accuracy and low latency in real-world test. Besides, we set a group of thresholds that can be easily changed for the duration and intensity of different motion definition, so that it can be adjusted according to the needs in future applications. This framework is robust with great potential for domestic caring in an ageless aging society.