Video-based handsign recognition for intuitive human-computer-interaction

Abstract

In this paper we present a video-based HCI-system for the recognition of 10 different hand postures in real time. Automatic removal of the forearmfromthe segmented hand object garantees a consistent input to the feature calculation step. An extensive comparison of three different approaches to feature extraction (Hu moments, eigencoefficients, fourier descriptors) was carried out. The combination of classifiers using different methods of feature description leads to a recognition rate of 99.5%, requiring only 15-17 ms per frame on a normal PC. The main contribution of this paper is the thourough evaluation, selection and combination of known steps. © Springer-Verlag Berlin Heidelberg 2002.