Fuzzy-genetic approach to identity verification using a handwritten signature

Abstract

Verification of the dynamic signature is an important issue of biometrics. There are many methods for the signature verification using dynamics of the signing process. Many of these methods are based on the so-called global features. In this paper we propose a new approach to the signature verification using global features. The proposed approach can be characterized as follows: (a) Classification of the signature is performed using a fuzzy-genetic system. (b) We select an individual set of features for each signer. (c) In the procedure of features selection we use a genetic algorithm with appropriately designed evaluation function. It works without access to the signatures called skilled forgeries (this is a major advantage of the proposed approach). (d) We determine weights of importance for evolutionarily selected features. (e) The weights are taken into account in the classification process. (f) An additional advantage of the proposed classifier is the possibility of its work interpretation and possibility of an analytical determination of its parameters without machine learning. In this paper we present the simulation results for the BioSecure signature database, distributed by the BioSecure Association.