Face Recognition Based on Maximizing Margin and Discriminant Locality Preserving Projection

摘要

As an efficient subspace learning algorithm in face recognition, locality preserving projections (LPP) is a linear projective map that arise by solving a variational problem that optimally preserves the neighborhood structure of data set. Though local neighborhood information held, it doesn't fully take discriminant classified information into account. Combined with maximum margin criterion (MMC), a new method called maximizing margin and discriminant locality preserving projections (MMDLPP) is presented in this paper to try to find the subspace that best discriminates different face classes and preserving the intrinsic relations of the local neighborhood in the same face class according to prior class-label information. The proposed method is compared with PC A as well as LPP. Experimental results on ORL, Yale, and YaleB face database convince us that the proposed method MMDLPP provides a better representation of the class information and achieves better recognition accuracy.