Offline Signature Verification Using Radial Basis Function with Selected Feature Sets

Abstract

This paper presents evaluation results of support vector machine (SVM) classifiers with radial basis function (RBF) kernel in offline signature verification. We have used two data sets of offline signatures and extracted 15 (fifteen) features from each signature sample of the data sets. The best feature subsets of the data sets were selected using filter and wrapper methods. For both the data sets, SVM classifiers with RBF kernel were designed with every selected feature sets individually. Classifiers were optimized, and their performances were evaluated using 10-fold cross-validation. Another classifier was designed using both the data sets combined to test the generalizability of the classifier across two different signatures.