Regularized discriminant analysis method based on virtual samples
Jing Xiao-yua · Microcomputer Information · 2010
The small sample size(SSS) problem will cause the singularity and instability of the per class covariance matrices. This paper based on specimen reconstruct uses the eigenvectors of covariance matrix to produce virtual training samples to overcome singularity of the per class covariance matrices. As a consequence, the classifier based on regularized discriminant analysis can be used directly in classification. The proposed method overcomes the problem that the multi-parameters of regularized discriminant analysis need optimizing. The experimental result demonstrates the effectiveness of the proposed approach in face recognition field.