A novel face description by 2DFLDA supervised projection gradient non-negative matrix factorization

作者：Gao Tao^*; Ma Xiang

来源：Journal of Computational Information Systems, 2011, 7(14): 5307-5313.

• 2DFLDA
• Fisher linear discriminant analysis
• Fisher linear discriminants
• Nonnegative matrix factorization
• ORL face database
• PGNMF
• RBF Network
• Recognition performance
• Small samples
• Yale face database

With analysis of Projection gradient non-negative matrix factorization (NMF) and two-dimensional Fisher linear discriminant analysis, and in view of the existence of the NMF Unsupervised Learning and the small sample problem, a novel feature extraction based on two-dimensional Fisher linear discriminant supervised projection gradient non-negative matrix factorization (SPGNMF) is proposed. The method is first to employ projection gradient non-negative matrix factorization to samples, and then the dimension of result image is not reduced directly by PCA. The subspace is constructed by vectors of mainly reflection difference between class, which can resolve small sample problem and get better discriminant feature. The sample is projected on subspace, lastly the coefficient of projection is classified by RBF. The Experimental results on the ORL face database and YALE face database show that the proposed method is feasible and higher recognition performance.

• 出版日期2011