Handwritten Signature Recognition Using Deep Learning

In an increasingly digital world, handwritten signatures remain a prevalent means of authentication for various legal and financial transactions. However, the rise of forgery and signature fraud has become a significant concern, necessitating robust automated solutions for signature verification. This research presents a novel approach for distinguishing between genuine and fraudulent handwritten signatures using Convolutional Neural Networks (CNN) and MobileNet, two popular deep learning architectures.

The primary goal of this study is to develop an accurate and efficient system capable of automated signature verification. To achieve this, a large dataset of both genuine and forged signatures is collected and preprocessed to create a suitable training dataset. The CNN and MobileNet architectures are then employed to extract essential features from the signature images, enabling the system to differentiate between authentic and fraudulent signatures effectively.

The research investigates various aspects of the proposed approach, including data preprocessing, model architecture selection, hyperparameter tuning, and performance evaluation. A comprehensive evaluation is conducted using well-established metrics such as accuracy, precision, and to assess the system's effectiveness in detecting signature fraud and classifying genuine signatures.

KEYWORDS: deep learning, CNN, Mobilenet, image processing, Handwritten signature, Fraud Detection, Signature Verification