A Hybrid Approach for Robust Signature Verification
P Saikumar, P. Ushashree, Archana Naik, Pavan Kalyan, Nissi · 2025
Signature verification is an important component in various security and authentication systems. One of the major challenges in such a domain is the capability to extract distinguishing features that accurately differentiate between real and forged signatures. This work proposes a hybrid feature extraction approach that combines the strengths of both Convolutional Neural Networks (CNN) and Histogram of Oriented Gradients (HOG) to improve signature representation. A decision tree-based feature selection method is used to refine the extracted features, which aims to improve the model performance for classifying between the real and forged signatures by taking the discriminative features. The approach was extended into a weighted ensemble classification model that integrates multiple classifiers. The outputs from these classifiers are aggregated using a weighted scheme, enabling robust final decision-making. This proposed hybrid method demonstrates a better and more robust model to verify the signature in practical applications.