An off-line signature verification system

Abstract

Signatures continue to be an important biometric because it remains widely used as a means of personal verification and therefore an automatic verification system is needed. In this paper we present an off-line signature verification and recognition system based on a combination of features extracted such as global features, mask features and grid features. The system is trained using a database of signatures. For each person, a centroid feature vector is obtained from a set of his/her genuine samples using the features that were extracted. The centroid signature is then used as a template which is used to verify a claimed signature. To obtain a satisfactory measure of similarity between our template signature and the claimed signature, we use the Euclidean distance in the feature space. The results were very promising and a success rate of 84.1% was achieved using a localized threshold.