Signatures Verification using CNN and HOG including Voting Classifier
B. Venkata Sivaiah, D. Vyshnavi, B. Mamatha, Mamilla Harish, A. Sathish Kumar, N. Siva, Ashok Kumar Patel · Advances in computer science research · 2024
This study suggests a unique hybrid feature extraction technique that expands the possibilities of Manual signature authentication systems.This method efficiently finds important characteristics in signature photos by combining Convolutional Neural Network (CNN) and Histogram of Oriented Gradients (HOG) approaches with a Decision Tree-based feature selection algorithm.Three classifiers were used in the evaluation: Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Long Short-Term Memory (LSTM).All three classifiers showed excellent accuracy in differentiating between genuine and fake signatures.Furthermore, a Voting Classifier (RF + DT) in the feature extraction process lead to an unparalleled 100% accuracy on testing datasets.This novel hybrid technique not only outperforms the findings of the original research but also demonstrates the resilience and adaptability of the suggested methodology, resulting in notable advancements in the performance of Manual signature authentication systems, especially against proficient forgeries.