Facial identification using transformed domain by SVM

Abstract

In this paper, we propose a facial biometric identification system, using transformed domains such as discrete cosine transforms (DCT), fast Fourier transforms (FFT) and discrete wavelets transforms (DWT). As novel classifier system has been utilized a support vector machines (SVM) (Burges, 1998), and in addition, the way of computing transformed coefficients. With the above system, it has been obtained a simple and robust system with good results, upper to 98% with the ORL face database (400 images) from Olivetti Research Laboratory.