Combination of Multi-class SVM and Multi-class NDA for Face Recognition

In this paper we propose a new framework for multi-class face recognition based on combination of support vector machine (SVM) and non-parametric discriminant analysis (NDA). SVM fully describes the decision surface by incorporating local information in the linear space. On the other hand, NDA is a non-parametric improvement over linear discriminant analysis that traditionally suffered from a fundamental limitation originating from the parametric nature of scatter matrices; however NDA by formulating the new form of scatter matrix in LDA detects the dominant normal directions to the decision plane. For our extension, we firstly describe the classification on multi-class datasets and then we propose a new formulation by combining multi-class SVM and multi-class NDA.