Offline Handwritten Signature Verification using CNN inspired by Inception V1 Architecture

In the field of behavioral biometric, signature verification is most referenced procedure for authentication of a person. A signature is considered to be the "seal of approval" for verifying the approval of a user and remains the most preferred means of authentication. This verification system mainly aims at verifying the discriminating the forged signature (forged by an imposter) from the genuine signatures. In this paper, Convolutional Neural Networks (CNN) have been utilized to learn features from the pre-processed genuine signatures and forged signatures. The CNN used is inspired by Inception V1 architecture(GoogleNet). The architecture uses the concept of having different filters on same level so that the network would be wider instead of deeper. In this paper, the proposed model is tested on few publicly available datasets such as CEDAR, BH-Sig260 signature corpus, and UTSig.