Multiple classifier for on-line signature verification system

Abstract

With the increase of advance development in security technology, many major corporations and governments start employing modern techniques to identify the identity of the individual. These include the adoption of a system such as on-line signature and handwriting verification application for banking systems, public sectors, as well as for documents and checks. To achieve better solutions, multimodal biometric system needs to be employed since this system exploits more than one psychological or behavioral at verification process. This work presents a signature verification system as behavioral system to ensure that the currency authentication is preserved by validating the genuine signature. This study developed signatures by applying multiple classification techniques. These include Artificial Neural Network (ANN), Support Vector Machine (SVM) and pearson correlation. These techniques are combined with fusion techniques, i.e., ordinal structure module of fuzzy and Or gate to determine the signature either it is real or forge. The average of the values we have it after applying multiple classification techniques is calculated, and the results are compared with the pre-defined threshold prior to decision making of either the signature is genuine or not. After collect many samples and calculate the final result we calculate the error rate for FRR and FAR to compare it with previous study. After calculated the error rate we found 2% for False Rejection Rate (FRR) and 0% for False Acceptance Rate (FAR), so the result for these study it’s better than previous one.