Data-Dependent Kernel Discriminant Analysis for Feature Extraction and Classification

Subspace analysis is an effective technique for feature extraction, which aims at finding a low-dimensional space of high-dimensional data. In this paper, a novel subspace analysis method based on data-dependent kernel discriminant analysis (DDKDA) is proposed for dimension reduction. The procedure of DDKDA contains two stages: one is to find the optimal combination coefficients by solving a constrained optimization function which transformed to an eigenvalue problem; other is to implement KDA under the optimal data-dependent kernel with Fisher criterion. DDKDA is more adaptive to the input data than KDA owing to the optimization of projection from input space to feature space with the data-dependent kernel, which enhances the performance of KDA. Experiments on the ORL and Yale face databases demonstrate the good performance of the proposed algorithm