Intrapersonal Parameter Optimization for Offline Handwritten Signature Augmentation

Usually, in a real-world scenario, few signature samples are available to train an automatic verification system (ASVS). However, such systems do indeed need lot of signatures achieve acceptable performance. Neuromotor duplication methods and feature space augmentation may be used meet the for increase number samples. Such techniques manually or empirically define set parameters introduce degree writer variability. Therefore, present study, method automatically model most common variability traits is proposed. The generate offline image ASVS. We also alternative approach evaluate quality considering their vectors. evaluated performance ASVS with generated using three well-known datasets: GPDS, MCYT-75, CEDAR. In GPDS-300, when SVM classifier was trained one genuine per duplicates space, Equal Error Rate (EER) decreased from 5.71% 1.08%. Under same conditions, EER 1.04% technique. verified that generates reproduces different datasets.