Signature recognition using majority voting

Abstract

Signature recognition is a highly dynamic study in the machine learning community. In this assignment, we tackled an important application: signature verification using artificial neural networks (ANN) and boosting. A ssignature is one of the most well-known personal characteristics for verification. It is straightforward, and suitable for users, official organizations, and courts. This paper focuses on offline signature verification. Two datasets have been used: the NFI-offline dataset, and the UTSIG dataset. Numerous image processing methods are utilized to identify and confirm the autograph. To validate the identity author, we use what we call the hard-voting scheme (also known as majority voting), every feature extraction method votes for an author, and the majority that obtains the most votes wins. This used software is MATLAB in which the autograph is taken and submitted in an image format.