Hand recognition based on finger-contour and PSO

Abstract

Hand shape recognition method based on geometric features uses individual information limitedly and inadequately. To solve this problem, this paper proposes a hand shape recognition method based on contour features of fingers. Firstly, we separate the four fingers and use curve fitting method to position the axis of finger. Then the matched fingers are normalized by translation and rotational alignment, so we can conduct the matching of contour features. Finally, in order to further improve the recognition rate, particle swarm optimization (PSO for short) is used to optimize the cut-off coefficient and the weight values of different fingers. Experimental results show that the proposed method can locate hand more accurately and make full use of hand information. It can also avoid the influence of inaccurate feature points locating and unstable contour around finger valleys. The recognition rate can reach 94.78%.