Digital Signature Recognition Using Geometric Features, Histogram of Oriented Gradients, Hu Moments, and Support Vector Machine

Abstract

The increasing use of electronic documents has heightened the need for fast, accurate, and objective digital signature verification systems. This study proposes a digital signature recognition system by combining Geometric Features, Histogram of Oriented Gradients (HOG), and Hu Moment feature extraction with a Support Vector Machine (SVM) classifier using the Radial Basis Function (RBF) kernel. A dataset of 500 signature images from 50 individuals was divided into training, validation, and testing sets using an 80:10:10 ratio. The proposed workflow includes image preprocessing, feature extraction, feature vector construction, model training, and evaluation using a confusion matrix. Experimental results show that the combined feature extraction methods effectively represent both global and local signature characteristics. The proposed model correctly classified 46 of 50 testing samples, achieving 92.00% accuracy, 88.00% precision, 92.00% recall, and an 89.33% F1-score, demonstrating its effectiveness for automatic digital signature recognition and electronic document authentication.