Abstract: Based on the Neighborhood Preserving Embedding (NPE) algorithm, a novel dimensionality reduction method called ONPE is proposed. First, a function which reflects the locality preserving power of the projective vectors is defined. Then, with the neighborhood preserving function as the objective function and the orthogonal constrained conditions added to the original optimal problem, the iterative formulae for finding a set of orthogonal optimal projection vectors are deduced. Compared with the NPE algorithm, the orthogonal vectors have the better locality preserving power, and thus the stronger discriminant power can be obtained and the error rate reduced. Experimental results on the standard face databases illustrate that in comparison with the other dimensionality reduction methods, the lowest error rate of the new method can be reduced by 15％~20％ and can be achieved when the number of the selected features is comparatively small.

Key words: neighborhood preserving embedding method, orthogonal neighborhood preserving embedding, locality preserving power, face recognition