ABSTRACT
The fingernail plate has been found to possess significant biometric attributes. The present authors have attempted to make a novel performance analysis of the fingernail plate in this chapter, using three different pre-trained deep learning models. A database of 890 (5 each from 178 volunteers) hand-dorsal images has been considered. The Region of Interest, i.e., the fingernail plate, has been extracted from the index, middle, and ring fingers of all these images. Features have been extracted using three pre-trained deep learning models, viz. AlexNet, ResNet-18, and DenseNet-201. The individual performances of all the three nail plates in both verification and identification systems have been examined first. The results obtained from these unimodal systems show the merit of the fingernail plate as a potential biometric trait. Various multimodal setups have also been designed using the nail plates in biometric authentication. The results attained appreciate the efficacy of this work as the multimodal verification systems perform reasonably well. Also, identification accuracy as high as 99.438% has been achieved from the designed multimodal system.
