Refinement of an online signature verification system

Abstract

The dynamic signature becomes a major concern of many researchers in the field of biometric because of its ability to protect many systems from being accessed by unauthentic persons. Many researchers have used different techniques to build their systems in order to have very good results in term of efficiency and performance. This research is focused on developing a graphical user interface Gl 1 with using two different classifiers which are global and local features of the on line signature. A digital tablet is used in this research as an input device to the system to convert the dynamic signatures to some sort of raw data to be preprocessed and extracting features from it and become an input after being classified as global and local for the thresholds stage which are back propagation neural networks and time and length of the signatures and by selecting the proper thresholds for them to be finally combined into an AND operation for the final decision. In this research an investigation on the output of back propagation neural networks took a place for further fine-tuning of the system to have more satisfactory performance on one of its techniques used. Finally, the expected result can be seen through the (False Rejection Rate FRR & False Acceptance Rate FAR) These rates will indicate this software's performance in verifying signature, thus it has to be as low as possible.