RaCon: A gesture recognition approach via Doppler radar for intelligent human-robot interaction

Abstract

As an important entrance for human-robot interaction, the hand gesture recognition based on wireless sensor has received great attention in recent years. By recognizing fine-grained arm movements, remotely deployed collaborative robot could work more accurately to satisfy human demands. Existing approaches mostly use wearable sensors or wireless devices to recognize human movement, which is with strict position requirements. In this paper, we propose a robust gesture recognition method based on double Doppler radars. Specifically, we use two Doppler radars to collect two sources of Doppler signal of a gesture. Then 6 types of gestures with different angles between people and the radar were classified by employing an improved dynamic time warping (DTW) algorithm. Furthermore, we demonstrate the practicability of the proposed method by developing a cooperative robot control system and the average recognition accuracy is 96%.