In this paper，a novel learning algorithm called center based neighborhood embedding(CNE) is proposed to deal with face recognition. Unlike the classical methods such as local linear embedding(LLE) and local preserving projection(LPP)，CNE is a supervised linear dimensionality reduction method. It first computes centers of all sample classes. The input of the weight function between two samples was replaced by center based neighborhood(CN) distance. Then，the high dimensional data are embedded into a low dimensional space with preserving the CN geometric structure. The CNE approach is compared with principle component analysis(PCA)，linear discriminant analysis(LDA) and local preserving projection(LPP) on ORL，Yale and UMIST databases. Experiments demonstrate the proposed method is superior to other three methods in terms of both lower dimensional visualization and recognition accuracy.