Discriminant analysis based on nearest feature line

Abstract

A novel feature extraction algorithm based on nearest feature line is proposed in this paper. The proposed algorithm can extract the local discriminant features of the samples. The performance of the proposed algorithm is directly associated with the parameter, so we use two discriminant power criterions to adaptively determine the parameter. Some experiments are implemented to evaluate the proposed algorithm and the experimental results demonstrate the efficiency of the proposed algorithm. © Springer-Verlag 2013.