LAUSR

Abstract Face is an important characteristic for biometric identification and verification systems, as it holds the needed information about individual’s identity. However, it is the characteristic most impacted by aging process, and facial aging is the main effect which decreases significantly the performance of face recognition algorithms. The main idea of this study relies on the fact that aging affects facial components such as mouth, eyes and nose differently. Therefore, we suggest to consider face as an independent component set, and each facial component (eyes, mouth, and nose) will be processed separately, and we propose an effective component-based method for age-invariant face recognition using a Discriminant Correlation Analysis(DCA) as a feature-level fusion algorithm to combine Deep-based features computed from separated facial components, and a Support Vector Machine (SVM) as a classifier. To evaluate the proposed approach, a comprehensive experimental study is performed on a widely used face-aging database, namely FG-NET, also the proposed method has been compared with a Particle Swarm Optimization (PSO)-based score-level fusion method. Experimental results show that the proposed component-based system worked effectively for face recognition across age progression, as it achieves significant high mean accuracy rates.