SVM-DSmT combination for off-line signature verification

We propose in this work a signature verification system based on decision combination of off-line signatures for managing conflict provided by the SVM classifiers. The system is basically divided into three modules: i) Radon Transform-SVM, ii) Ridgelet Transform-SVM and iii) PCR5 combination rule based on the generalized belief functions of Dezert-Smarandache theory. The proposed framework allows combining the normalized SVM outputs and uses an estimation technique based on the dissonant model of Appriou to compute the belief assignments. Decision making is performed through likelihood ratio. Experiments are conducted on the well known CEDAR database using false rejection and false acceptance criteria. The obtained results show that the proposed combination framework improves the verification accuracy compared to individual SVM classifiers.