›› 2015, Vol. 28 ›› Issue (2): 51-.

• 论文 • 上一篇    下一篇

一种快速非负矩阵分解的人脸识别算法

郑颖   

  1. (西安电子科技大学 数学与统计学院,陕西 西安 710126)
  • 出版日期:2015-02-15 发布日期:2015-02-16
  • 作者简介:郑颖(1988—),女,硕士研究生。研究方向:人脸识别,模式识别。E-mail:zhengyin@163.com

A Rapid Non-negative Matrix Decomposition Method for Face Recognition

ZHENG Ying   

  1. (School of Mathematics and Statistics,Xidian University,Xi'an 710126,China)
  • Online:2015-02-15 Published:2015-02-16
  • About author:[1]Turk M,Pentland A.Eigenfaces for recognition[J].Journal of Cognitive Neurosci,1991,3(1):71-86. [2]Jolliffe I T.Principal component analysis[M].2nd Edition.Berlin:Springer,2002. [3]Yang J,Zhang D,Frangi A F,et al.Two-dimensional PCA:A new approach to appearance based face representation and recognition[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,2004,26(1):131-137. [4]刘维湘,郑南宁,游曲波.非负矩阵分解及其在模式识别中的应用[J].科学通报,2006,35(1):241-250. [5]严慧,金忠,杨静宇.非负二维主成分分析及其在人脸识别中的应用[J].模式识别与人工智能,2009,6(22):809-814. [6]高红娟,潘晨.基于非负矩阵分解的人脸识别算法的改进[J].计算机技术与发展,2007,17(11):63-66. [7]高红娟,潘晨.基于(2D)2NMF及其方法的改进[J].计算机应用,2007,27(7):1160-1166. [8]Yang J,Zhang D,Yong X.Two-dimensional discriminant transform for face recognition[J].Pattern Recognition,2005,28(7):1125-1129. [9]Ming L,Zong Y B.2D-LDA:A statistical linear discriminant analysis for image matrix[J].Pattern Recognition Letters,2005,26(12):527-532. [10]Kong H,Teoh E K.Generalized 2D fisher discriminant analysis[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,2012,71(19):5244-5251.

摘要:

用于人脸识别的非负矩阵分解算法,虽可提高图像识别率,但因其是通过迭代方法同时计算出基矩阵和系数矩阵,故当迭代次数较多时,计算过程耗时长。文中将二维线性判别分析方法与非负矩阵分解方法融合,提出了一种快速的双边二维非负矩阵分解算法。通过在AR、Yale人脸数据库上的实验结果显示,较二维双边非负矩阵分解算法,文中算法不仅使得训练时间大幅减少,而且识别率也有所提高。

关键词: 非负矩阵分解, 特征提取, 人脸识别

Abstract:

Face recognition algorithms through non-negative matrix factorization can increase image recognition rat,but it is an iterative algorithm which must simultaneously calculates the base matrix and the coefficient matrix,leading to high computational complexity with many iterations.This paper introduces the 2-dimensional linear discriminant analysis into the non-negative matrix factorization algorithm,and proposes a fast two-dimensional bilateral non-negative matrix factorization algorithm.Experiment results on the AR and the Yale face database show that the proposed algorithm has higher recognition performance as well as a much faster speed than the two-dimensional bilateral non-negative matrix factorization algorithm.

Key words: non negative matrix factorization;feature extraction;face recognition

中图分类号: 

  • TP391.41