Research on Features Extraction from Frame Image Sequences

Abstract

Human hand gesture features extraction from the frame image sequences is one of the key jobs in the process of human hand tracking based upon particle filtering (PF), because it provides PF with key data for computing the weight value of a particle. In order to satisfy the need of real time human hand tracking, a novel features extraction approach, aimed at optimization of processing speed, is put forward in this paper. First of all, the hand contour is approximately described by a polygon with concave and protruding, and the relationship between hand gesture polygon and its bounding box is studied. Secondly, a hand gesture contour algorithm(HGCA) is proposed to obtain the hand gesture polygon. Then, based on the HGCA, the approaches to gain the several main feature points are presented, including fingertips, roots of fingers, joints and the intersection of knuckle on different fingers. Finally, some of the comparison experimental results are presented. The main advantages of human hand features extraction put forward in this paper consist in its controllable precision and low time complexity.