Verification System for Handwritten Signatures with Modular Neural Networks

Abstract

Handwritten signature is considered as one of the primary biometric processes for human verification in various applications including banking and legal documentations. In general, the handwritten signatures are verified with respect to the pressure, direction and speed followed on a plain document. However, the traditional methods of verification are less accurate and time consuming. The proposed work aims to develop a deep learning -based approach for handwritten signature verification process through a Modular Neural Network algorithm. The work utilized the handwritten signatures dataset downloaded from the kaggle website that consists of original and forged signatures of 30 individuals. The work also included a set of 20 individual signatures for improving the sample count on training and verification process.