The Application Of Local Nuclear Non-negative Matrix Factorization Algortihm To Face Recognition

Face recognition is an interdisciplinary frontier challenging issues, the main contents of its research is how to make computers with the ability of identification. Face recognition technology involved in a lot, which is a key feature extraction and classification methods, this paper focus on this research.Non-negative matrix factorization (NMF) algorithm and local non-negative matrix factorization (LNMF) algorithm is based on the local features of the feature extraction and have been successfully used for face recognition. But in the face recognition, NMF algorithm has a low recognition rate, LNMF algorithm improved the recognition rate in a way, but its price is to increase the number of iterations. In addition, the two algorithms have not been very good to resolve the issue of non-linear separable. This paper combines nuclear methods and LNMF algorithm produce a nuclear local non-negative matrix factorization (KLNMF) algorithm, firstly through the nonlinear transform conversion the original space to high-dimensional space, making samples linearly separable, then LNMF algorithm extract facial features. Decision-making in the classification of the papers put forward its own classification rules, and the design based on the NMF subspace classifier.This paper, usage the iris data and ORL face database, analyzes performance of KLNMF algorithm compared with the NMF, LNMF algorithm. Experiments show KLNMF algorithm can effectively solve the nonlinear problem. It not only improved the recognition rate a large extent, but also avoided the increase in the number of iteration.