A Deep Learning-Based Multimodal Recognition Biometric System Using Finger Vein and Finger Knuckle Print
Mouhamed Laid Abimouloud, Khaled Bensid, Monji Kherallah · 2023
Secure personal identification remains a paramount challenge in contemporary society, necessitating robust and reliable identity verification mechanisms. Biometric identification systems offer a promising avenue, given their inherent secu-rity attributes. In particular, the finger knuckle print (FKP) and finger vein (FV) are emerging as potent hand biometric modalities. In this paper, we employed convolutional neural networks (CNNs), foundational structures in deep learning, owing to their unparalleled proficiency in image analysis. Leveraging transfer learning through pretrained CNN models, we evaluated a biometric system that employs both finger knuckle and finger vein modalities. This paper proposes utilising two distinct datasets comprising 11,736 images 7,920 of finger knuckles and 3,816 of finger veins. Further enhancing system efficacy, we implemented a multimodal approach by fusing unimodal scores, our suggested multimodal system EER performance in finger knuckle print (FKP) 0.000% and finger vein (FV) 0.00599%. Preliminary experimental outcomes underscore the promising potential of FKP and FV in biometric applications.