Writer identification using dynamic features

Abstract

Automatic handwriting identification is an interesting research topic in the community of biometrics. In this paper, we propose a new approach to writer identification based on dynamic features, namely the features associated with gestures and movements of the writers during writing. The feature vectors defined here are mainly extracted from the writing pressure, velocity and pen inclination, whose components are selected according to respective reliability. Experimental results demonstrate that the proposed algorithm performs well, especially in tackling the degraded documents with little loss of accuracy. © Springer-Verlag Berlin Heidelberg 2004.