Automatic Shape Recognition of Hand Gestures Using an Edge-tracing Vision System

This paper describes a working computer visionsystem and software algorithms that can automatically recognizedifferent types of hand gestures. The software performs theimage capture, analysis and shape identification for threedifferent kinds of hand gestures, namely, the ‘rock’, ‘paper’ and‘scissors’ hand gestures used in a popular child’s game. Thesoftware is comprised of three main elements, namely: (1) Amethod for collecting data to store normalized or ‘template’(standard) images of known shapes or known hand gestures; (2)An edge detection algorithm; and (3) A method to compare animage with normalized or ‘template’ images. This method ofhand gesture recognition could be adapted for use in manypractical applications relating to the control of computers (e.g.3D graphics, controlling games, real-time simulations), robotmanipulators, mobile robots and machinery.