Convolutional neural networks applied to the biometrics person recognition

Abstract

Person biometric recognition has been widely employed in the last decades aiming a better security level in a wide sort of applications. Amongst the available techniques, one of the most reliable is the fingerprint, which is commonly used for forensic detectives, since it can be acquired without an individual consent in a crime scene, for instance. In this work, we propose to study the viability of using Convolutional Neural Networks (CNNs) in fingerprint images aiming a person automatic identification, since a few works (or even none) have been performed in this context using such deep learning approach. This work also aims an internship through "Bolsa de Estágio e Pesquisa no Exterior" - BEPE program together with Harvard University, in which a research group has been working with parameter optimization of CNNs.