Uncorrelated enhanced diversity fisher discriminant analysis for face recognition

Abstract

In this paper, based on the enhanced fisher discriminant criterion (EFDC), a new feature extraction method called uncorrelated enhanced diversity fisher discriminant analysis (UEDFDA) is proposed for face recognition. UEDFDA defines the parameterless diversity weighted matrix by taking both the class label information and the local structure into account. Thus UEDFDA can preserve the local diversity structure of the data without setting any parameters. Moreover, UEDFDA is able to extract the uncorrelated discriminant vectors in the feature space and overcomes the small sample size problem, which is desirable for face recognition. Experimental results on the face databases show the feasibility and validity of the proposed method.