부산대학교 도서관

This paper presents an automatic 3D dynamic hand gesture recognition algorithm relying on both intensity and depth information provided by a Kinect camera. Gesture classification consists of a decision tree constructed on six parameters delivered by the Hough transform of projected 3D points. The Hough transform is originally applied, for the first time, on the projected gesture trajectories to obtain a reliable decision. The experimental data obtained from 300 video sequences with different subjects validate the proposed recognition method.