A novel architecture to verify offline hand-written signatures using convolutional neural network

Hand-written signatures are marked on documents to establish legally binding evidence of identity and intent. However, they are prone to forgery, and the design of an accurate feature extractor to distinguish between highly-skilled forgeries and genuine signatures is a challenging task. In this paper, we propose a Convolution Neural Network (CNN) architecture for Signature Verification (SV). The algorithm is trained using two signatures, genuine and forged. Then the SV module performs a classification task to determine if any two signatures are of the same individual or not. The simulation results show that the proposed method can achieve 27% (relatively) better results than the benchmark scheme. The paper also integrated different data augmentation techniques for the signature data, which further improved the efficiency of the proposed method by 14% (relative).