Face Recognition Based on Kernel Schur-Orthogonal Neighborhood Preserving Discriminant Embedding

Abstract

In order to recognize faces more accurately, this paper proposes a new manifold learning algorithm named Kernel Schur-Orthogonal Neighborhood Preserving Discriminant Embedding (KSONPDE) which puts the vector orthogonal and kernel mapping into the Neighborhood Preserving Discriminant Embedding (NPDE). The algorithm extracts nonlinear information from face image by kernel method, mapping it into a high-dimensional space and finding optimal projection vector by schur-orthogonal when solving eigenvalues in order to extract the face features from the structure of nonlinear local area. The experiment on the ORL and Yale face database demonstrates effectiveness of the proposed method.