A Dorsal Hand Vein Recognition System based on Various Machine and Deep Learning Classification Techniques

Dorsal vein biometrics is a widely used type of biometric identification due to the distinctiveness of the vein patterns, even between the right and left hands of the same person. In this study, a recognition system for human dorsal hand veins was proposed, utilizing various machine and deep learning methods such as KNN, SVM, DT, RF, and CNN on two hand dorsal vein datasets. The system demonstrated commendable performance on both datasets, achieving accuracy rates of 99.3%, 96%, 91.4%, 97.7%, and 99.8% for KNN, SVM, DT, RF, and CNN, respectively, even with the augmented dataset. However, the KNN and CNN models were found to be the most effective, producing high F1-scores, accuracy, and recall.