Offline Signature Verification using Spatial Domain Feature Sets and Support Vector Machine
K N Pushpalatha, Arvind Kumar Gautam · 2014
Abstract—Biometrics has become an accepted form of identification of an individual in the modern world. Offline signature verification system has become one of the biometrics legally accepted form of biometric identification techniques. The transform domain methods often generate larger feature vectors which increase the computation time. An effort is made in this paper to reduce the dimension of feature vector dimension so as to reduce the computation time while still maintaining a higher recognition rate. The strong feature set consists of Concentric Square with five unique Geometric features and eight Camastra features being computed from each square. In this work it has been observed that Support Vector Machine (SVM) as a classifier provides a better performance. The experimental results on GPDS-960 database images have produced improved False Acceptance Rate (FAR), False Rejection Rate (FRR) and Total Success Rate (TSR) as compared to other algorithms. Keywords—Camastra features, Concentric Squares, Geometric features, HMM, SVM.