Signature Verification Using Deep Learning: An Empirical Study

A critical step in biometric recognition is signature verification, which has applications in forensic investigations, legal papers, and financial transactions. In this study, we offer an autonomous feature extraction and learning method for convolutional neural networks (CNNs) that can automatically learn and extract pertinent features from input pictures for classification. The three key steps of our suggested methodology are pre-processing, feature extraction, and classification. To eliminate noise and distortions, we normalize the signature image and remove the backdrop during the pre-processing step. We train a CNN model to extract deep features from the processed signature picture during the feature extraction step. Finally, for classification, we use a fully linked layer. We test the proposed system using the CEDAR and CEDAR signature databases, which are both freely accessible. In comparison to state-of-the-art approaches, the experimental findings show that our suggested method is successful in reaching high accuracy 89.2%.