Orthogonal Discriminant Linear Local Tangent Space Alignment For Face Recognition

摘要

In this paper, a novel linear subspace learning algorithm called orthogonal discriminant linear local tangent space alignment (ODLLTSA) is proposed. Derived from linear local tangent space alignment (LLTSA), ODLLTSA not only inherits the advantages of LLTSA which uses linear local tangent space as a representation of the local geometry to preserve the local structure, but also makes full use of class information to improve recognition power, solves the optimal subspace by spectral regression, and then orthogonalizes the subspace. Experimental results on standard face databases demonstrate the effectiveness of the proposed algorithm.