Signature Verification based on Reduced Size Fast Correlation

Abstract

In this dissertation work, performed an far-reaching experimental study to minimize time of recognition and verification and increase the great accuracy of signature sample which extract some features from signature sample in their training phase. In previous research on signature verification was based on similarity training of machine used in neural network has shown better accuracy than other techniques but somewhere neural network is complex in operation for example different algorithms for supervised learning. So in this work feature extraction and then it's database has been prepared which takes less time than supervised training in NN is used and for verification, correlation is used to match features of signature with the database which is also easier in operation and faster in processing then BP algorithms in NN. In this system more than 400 signature samples as used for recognition, and it gives far-accuracy than NN system.