Offline Signature Verification & Recognition Using Angle Based Feature Extraction & Neural Network Classifier

The concept of the signature is wide used as a means of individual verification emphasizes the need for an automatic verification system because of the individual aspect effect of being easily misused by those who would pretend the Identification of an individual. A great deal of work has been done in the area of off-line signature verification over the past few decades. Verification can be performed either Offline or Online based on the application. In this paper, we present a method for Offline Verification of signatures by calculating the angle curvature of each direction in signature samples. Here we have taken 8bins and calculated in all directions and extracted the features and stored in templates. The features that are used are Area, Centre of gravity, Eccentricity and Skewness along with angle based features. Earlier the features are extracted, pre-processed a scanned image to isolate the signature part and to remove any false noise present. The system is initially trained using a database of signatures. In our experimentation MCYT signature database is used. An accuracy of 97.61 % with Feed forward Back Propagation is observed in identifying the test signatures.