Neural Network Based Signature Verification Model for Bank Cheques with Three Specimen Signatures
Subramaniam Dhandapani, TJPRC · International journal of computer science engineering and information technology research · 2017
Researches in offline signature verification use a large number of genuine and forged samples for the verification purpose, whereas real time banking system uses only a limited number of genuine samples.To go along with the reality, this research paper proposes a system to verify signatures on bank cheques, using only 3 signature samples.It does not require forged signatures by skillful forgers, as it may lead to more number of cheque frauds in future.The entire system modeled is divided into recognition and verification phase.Recognition phase is further subdivided into supervised and unsupervised methods.In the supervised recognition method, the account number of the customer is given as input and the corresponding signature is retrieved.But, in the unsupervised recognition method, the signature is directly given as input and the signature matching the features of the query signature are retrieved.So, the unsupervised input method has increased the system efficiency by reducing the processing time.Here, Feed forward Back propagation Neural Network is used for the classification purpose.So, the proposed verification model goes with the real time banking process for verification of signature on cheque and yields excellent results with an overall accuracy of 91.33%.Thus, this paper has successfully overcome the existing hurdles in offline signature verification, like requiring more number of samples and involving forging experts.It has also improved the efficiency of the system by the use of unsupervised recognition and verification method.