计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (17): 175-177.

• 图形、图像、模式识别 • 上一篇    下一篇

核非负矩阵因子及其在表情识别中的应用

黄 勇   

  1. 柳州铁道职业技术学院 电子工程系,广西 柳州 545007
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-06-11 发布日期:2011-06-11

Kernel non-negative matrix factorization and its application to expression recognition

HUANG Yong   

  1. Department of Electronic Engineering,Liuzhou Railway Vocational Technical College,Liuzhou,Guangxi 545007,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-06-11 Published:2011-06-11

摘要: 提出了一种基于核非负矩阵因子分解的人脸表情识别方法。该算法引入核函数并结合NMF进行表情特征提取,称之为PNMF、GNMF。与NMF等不同,PNMF、GNMF通过基于核的非线性映射可从原始表情数据中提取更多的有用信息,包括线性的和非线性的,尽可能地保留原始的表情信息。基于CED-WYU(1.0)和JAFFE两个表情数据库的识别结果表明,基于核的NMF特征提取方法能有效地提高识别率及效率。

关键词: 非负矩阵因子分解, 极核非负矩阵因子分解, 高斯核非负矩阵因子分解, 表情识别

Abstract: The facial expression recognition methods based on kernel non-negative matrix factorization are proposed.Unlike NMF,through using kernel induced nonlinear mapping,PNMF and GNMF can extract more useful expression features hidden in the orginal expression image,thus remain the original expression information as much as possible.Experimental result on CED-WYU(1.0) and JAFFE shows that it is an effective method for improving the recognition accuracy and effectiveness.

Key words: Non-negative Matrix Factorization(NMF), Polynomial kernel NMF(PNMF), Gaussian kernel NMF(GNMF), expression recognition