New features for authentication by on-line handwritten signatures

Abstract

Authentication by handwritten signature is the most accepted authentication system based on biometrics partly due to the fact that it is easy to be used and part of our cultural habits. We present some improvement for one of the most important problems in such systems: features construction and selection. Concerning the study of the signature's shape, we discuss the impact of the selected features on the global system architecture. We propose to extract information linked to the writer by computing new features based on fractal dimension of the on-line signature. In order to prove the interest of these features, several performance evaluation methods are presented. Finally, the results of tests using the combination of new features with classical ones, and a method of features selection using Genetic Algorithm are discussed. © Springer-Verlag Berlin Heidelberg 2004.