An optimization criterion for generalized marginal Fisher analysis on undersampled problems
International Journal of Automation and Computing
Nearest-neighbor classifier motivated marginal discriminant projections for face recognition
Frontiers of Computer Science in China
Hi-index | 0.00 |
A new face recognition algorithm, termed enhanced marginal fisher analysis (EMFA), is proposed in the paper. Different from MFA in which the construction of the interclass graph is based on the whole dataset, that is usually time-consuming, EMFA first find the nearest classes of each class using the mean vector of each class, then the marginal points can be directly selected from their nearest classes. Compared with the original MFA, the proposed method has a better efficiency for face recognition, and can avoid overfitting effectively. Experimental results on the ORL and FERET face databases show EMFA outperforms other methods.