Cancellable Face Biometrics Template Using AlexNet

Abstract

Biometric systems with traits like gesture, voice, fingerprint, palm print, handwritten signature, hand geometry, face, and iris have been utilized for authentication. Through these traits, face trait is considered as one of the strongest and an important biometric element. In this work, a presentation of a new cancellable algorithm of face image which is dependent on AlexNet and Winner-Takes-All (WTA) hash method has been proposed. AlexNet is a Convolutional Neural Networks (CNNs) that reached a state-of-the-art level of recognition precision compared to other conventional machine learning methods in terms of feature execution. WTA is used for similarity purposes, whereas random binary orthogonal matrices are applied to produce the projected features of vectors. Fundação Educacional Inaciana dataset and Georgia tech face dataset were used in evaluating the performance of the proposed algorithm. Experimental results illustrate the proposed algorithm has satisfactory execution performance in terms of Equal Error Rate. Thus the proposed algorithm can be used as an alternative method in security biometric implementation.