A new face recognition method, realized by the feature fusion technique based on Local Non-negative Sparse Coding (NNSC) and Local Non-negative Matrix Factorization (LNMF) algorithms, is proposed in this paper. NNSC and LNMF are both part-based representations of the multi-dimensional data, used widely and efficiently in image feature extraction and pattern recognition. Here, considered the high recognition rate, the weighting coefficient fusion method between features obtained by algorithms of NNSC and LNMF is discussed in the face recognition task. Using the distance classifier and the Radial Basis Probabilistic Neural Network (RBPNN) classifier, the recognition task is easily implemented on the ORL face database. Moreover, compared with any other algorithm of NNSC and LNMF, experimental results show that the feature fusion method is indeed efficient and applied in the face recognition.