Comparing the performance of principal component analysis and rbf network for face recognition using locally linear embedding

Among the many methods proposed in the literature for face recognition, those relying on face manifold have been explored with great interest in the last few years.In those methods theface images are initially subjected to dimensional reduction and then applied to a classifier.In this paper we evaluate and compare two novel approaches for face recognition to address thechallenging task of recognition using a fusion of nonlinear dimensional reduction;Locally Linear Embedding (LLE) integrated with Principal Component Analysis (LLEPCA) and LLE with RBF network (LLERBF).Extensive experiments using the CMU AMP Face EXpression Database and JAFFE databases indicate that PCA is faster than RBF in regard to computation time when we measured the CPU time.The difference was almost 1.2 second when we run both algorithms on the same data.