Handwritten Signature Recognition Using Siamese Neural Network

This research proposes a novel approach to addressing the issue of signature authenticity identification by implementing Deep Metric Learning (DML). The aim of this study is to develop a DML model with high accuracy in training and testing processes that can be integrated with a mobile application for signature authenticity verification. The data used in this research are signature images from 90 individuals, divided into two categories: genuine signatures and forged signatures. Testing results show that the developed model has an accuracy of 85.7<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, with high precision for the forged signature class at 95.9% and recall for the genuine signature class at 94.9%. The conclusion of this research is that the developed DML model is reliable for signature verification, although further development is needed to improve the model's performance, especially in detecting forged signatures. Suggestions for future research include increasing the amount and variety of training data, integrating the model with a mobile application, and testing the model under various conditions and on different devices.