Discriminant maximum margin projections for face recognition

In this paper, we propose a novel dimensionality reduction algorithm called discriminant maximum margin projections (DMMP) for face recognition. By discovering both geometrical and discriminant structures of the data points, DMMP aims at finding a subspace that optimally preserves the local neighborhood information of the data set, as well as maximizes the margin between data points from different classes at each local area. Moreover, DMMP utilizes a equilibrium parameter to adjust the significance of the locality preserving property and margin distances of the data points. Finally, with the experiments used face recognition data sets, such as the ORL, Yale, and FERET face databases, the results prove that DMMP can attain a better effectiveness than most other advanced approaches.