OSVNet: Convolutional Siamese Network for Writer Independent Online\n Signature Verification

Online signature verification (OSV) is one of the most challenging tasks in\nwriter identification and digital forensics. Owing to the large\nintra-individual variability, there is a critical requirement to accurately\nlearn the intra-personal variations of the signature to achieve higher\nclassification accuracy. To achieve this, in this paper, we propose an OSV\nframework based on deep convolutional Siamese network (DCSN). DCSN\nautomatically extracts robust feature descriptions based on metric-based loss\nfunction which decreases intra-writer variability (Genuine-Genuine) and\nincreases inter-individual variability (Genuine-Forgery) and directs the DCSN\nfor effective discriminative representation learning for online signatures and\nextend it for one shot learning framework. Comprehensive experimentation\nconducted on three widely accepted benchmark datasets MCYT-100 (DB1), MCYT-330\n(DB2) and SVC-2004-Task2 demonstrate the capability of our framework to\ndistinguish the genuine and forgery samples. Experimental results confirm the\nefficiency of deep convolutional Siamese network based OSV by achieving a lower\nerror rate as compared to many recent and state-of-the art OSV techniques.\n