基于分块非负矩阵分解人脸识别增量学习

被引:6
作者
潘彬彬
陈文胜
徐晨
机构
[1] 深圳大学数学与计算科学学院智能计算科学研究所
基金
广东省自然科学基金;
关键词
非负矩阵分解; 局部特征提取; 人脸识别; 增量学习;
D O I
暂无
中图分类号
TP391.41 [];
学科分类号
080203 ;
摘要
非负矩阵分解(NMF)算法可以提取图像的局部特征,然而NMF算法有两个主要缺点:a)当矩阵维数较大时,NMF算法非常耗时;b)当增加新的训练样本或类别时,NMF算法必须进行重复学习。为克服NMF算法这些缺点,提出了一种新的分块NMF算法(BNMF)。特别地,该方法还可用于增量学习。通过在FERET和CMUPIE人脸数据库上进行实验,结果表明该算法均优于NMF和PCA算法。
引用
收藏
页码:117 / 120
页数:4
相关论文
共 13 条
[1]  
Learning the Parts of Objects by Non-negative Matrix Factorization. Lee D D,Seung H S. Nature . 1999
[2]  
Non-Negative Sparse Coding. P O Hoyer. NeuralNetworks for Signal Processing . 2002
[3]  
Projected gradient methods for nonnegative matrix factoriza-tion. LIN C J. Neural Computation . 2007
[4]  
A novel discriminant non-ne-gative matrix factorization algorithm with applications to facial imagecharacterization problems. KOTSIA I,ZAFEIRIOU S,PATAS I. IEEE Trans on Information Foren-sics and Security . 2007
[5]  
Non-negative matrix factorization with sparseness constraints. Hoyer,P. O. Journal of Machine Learning Research . 2004
[6]  
Fisher non-negative matrixfactorization for learning local features. Wang Y,,Jiar Y,Hu C,et al. Proceedingsof the Asian Conference on Computer Vision(ACCV) . 2004
[7]  
Algorithms for non2negative matrix factorization. Lee D D,Seung H S. Proceedings of Neural Information Processing Systems . 2000
[8]  
Eigenfaces for recognition. Turk M,,Pentland A. Journal of Cogni-tive Neuroscience . 1991
[9]  
A new sparse image representation algorithm ap-plied to facial expression recognition. BUCIUI,PITAS I. Proc of the 14th IEEEWorkshop on Machine Learning for Signal Processing . 2004
[10]  
A modified non-negative matrix factorization algorithm for face recognition. XUE Yun,TONG C S,CHEN Wen-sheng,et al. Procof the 18th International Conference on Pattern Recognition . 2006