Planar multi-classifier modelling-NN/SVM: Application to off-line handwritten signature verification

Abstract

We present in this paper a new approach to modeling an image in a pattern recognition context. It relates to a generalization of the planar Markov modeling, which we shall call PlanarMulti-ClassifierModeling (PMCM). Thesemodels allowexploring various classifier types, among others, Markov ones, to model delimited bands of an image associated with a given pattern. Determining the different parameters of the proposed planar model ismade explicit. Various PMCMarchitectures are proposed; in this paper we present the PMCM-NN/SVM architecture. The validation has been performed in a security context based on off-line handwritten signature verification. The different experiments, carried out on two public databases: the SID-Signature database and the GPDS-160 Signature database, have led to promising results.