Stepwise nearest neighbor discriminant analysis

摘要: Linear Discriminant Analysis (LDA) is a popular feature extraction technique in statistical pattern recognition. However, it often suffers from the small sample size problem when dealing with high dimensional data. Moreover, while LDA guaranteed to find best directions each class has Gaussian density common covariance matrix, can fail if densities are more general. In this paper, new nonparametric method, stepwise nearest neighbor discriminant analysis(SNNDA), proposed point of view classification. SNNDA finds important without assuming belong any particular parametric family. It does not depend on nonsingularity within-class scatter matrix either. Our experimental results demonstrate that outperforms existing variant methods and other state-of-art face recognition approaches three datasets ATT FERET databases.