Facial Recognition: A Combined Approach Utilizing CICA, PCA, and FFT

Abstract

Facial recognition (visual Biometric) is an individual identification system based on analyzing facial features. It is used to authenticate or identify persons by comparing a facial image with images in facial database or any other digital image sources. The process of facial recognition (FR) should be accurate regardless of image affecting factors (translation, expression, illumination, etc.), which is still a challenging problem. In this research, a combined Facial recognition technique is suggested utilizing principle component analysis (PCA), Constrained Independent Component Analysis (CICA), and Fast Fourier Transform (FFT). FFT is used to overcome the problems of translation, and slight illumination sensitivity. The paper aims, at first, to study the effect of combining PCA with CICA for facial recognition. The next step is to evaluate the systems performance when transforming facial images (using FFT) before applying PCA and CICA. The experimental results (conducted on ORL & Yale databases) indicate that the accuracy is highly improved when FFT is applied.