Nearest neighbour line nonparametric discriminant analysis for feature extraction
A new feature extraction method, called nearest neighbour line nonparametric discriminant analysis (NNL-NDA), is proposed. The previous nonparametric discriminant analysis methods only use point-to-point distance to measure the class difference. In NNL-NDA, point-to-line distance with nearest neighbour line (NNL) theory is adopted, and thereby more intrinsic structure information of training samples is preserved in the feature space. NNL-NDA does not assume that the class densities belong to any particular parametric family nor encounter the singularity difficulty of the within-class scatter matrix. Experimental results on ORL face database demonstrate the effectiveness of the proposed method.
