Off-line Signature Verification with concentric squares and slope based features using support vector machines

An Off-line Signature Verification System (OSVS) with a novel feature extraction procedure has been described. Fusion of concentric squares having geometric features, zone based slope as well as slope angle have been considered as input patterns. The strong feature set thus obtained makes the OSVS accurate. Verification was performed by using Support Vector Machine (SVM) technique with different kernels. Empirically, Radial Basis Function (RBF) based SVM model exhibited the best results as compared to that based on linear and polynomial kernels. That is, the system attained False Acceptance Rate as 1.25% and False Rejection Rate as 1.66%.