A new classification method for PCA-based face recognition

Abstract

This paper introduces a novel pattern classification approach called l 1 norm nearest neighbor convex hull (l1 NNCH) approach and applies it for PCA-based face classification. In l1 NNCH, l 1 norm distance from a query to a convex hull of a class is defined as the similarity of nearest neighbor rule. Principle component analysis (PCA), as an efficient technology for extracting feature, is applied to extract features of faces in this paper. Experimental results on the ORL and NJUST603 face databases show that l1NNCH combined with PCA has a good performance for face recognition. © 2009 IEEE.