Adaptive MobileNet Framework for Human Signature Recognition

In an era of more digital interaction, authenticating persons through handwritten signatures is crucial. This study presents an Adaptive MobileNet Framework for accurate hand-written signature recognition, integrating MobileNet's lightweight architecture with adaptive methodologies to enhance performance. The framework improves feature extraction and utilizes transfer learning to increase accuracy across various signature styles. We rigorously assess the model's efficacy utilizing an extensive dataset of authentic signatures from various sources. Experimental findings indicate that the Adaptive MobileNet Framework far surpasses conventional recognition systems, with an accuracy of 91.5%, precision of 90.8%, and recall of 91.2%. The framework is optimized for optimal computing resource utilization, making it appropriate for mobile and embedded devices, hence facilitating real-time signature verification.