A framework for offline signature verification system: Best features selection approach

Biometric verification is a method of identifying the persons by their individualities or traits. Signature verification is the most generally used biometric to maintain human privacy. It is used in many areas as banking, access control, e-business etc. and equally important in financial transactions. Research has progressed greatly in the area of signature verification but still, it is hard to discriminate between genuine signatures and skilled forgeries. Based on the idea of best features selection, a novel technique is introduced in this article foran offline verification system. The presented system consists of four major steps: preprocessing, features extraction, features selection, and feature verification. Global features in the proposed work comprise of aspect ratio, the area of signature, pure width, pure height and normalized actual signature height. Local features consist of signature centroid, slope, angle, and distance. In features selection component, a genetic algorithm is utilized to find appropriate features set which are later on given to support vector machine for verification. For experimental analysis, the selected datasets are CEDAR, MCYT and GPDS synthetic. The performance of proposed algorithm is based on three accuracy measures as FAR, FRR and AER. (C) 2018 Elsevier B.V. All rights reserved.