Orthogonal manifold preserving projections
Luo Da-yong · Journal of Central South University(Science and Technology) · 2011
In order to overcome the shortcomings of locality preserving projections(LPP),a new feature extraction methods orthogonal manifold preserving projections(OMPP) was proposed.The non-locality information of data is incorporates to the objective function,in contrast to LPP,which preserves the local and global structure of the data manifold.Gram-Schmidt orthogonalization was used to determine basic orthogonal vectors,which solves the non-orthogonal problem of LPP.Experiments on ORL and Yale face databases were performed to test and evaluate the proposed algorithm,and the results show that the proposed method is effective.