计算机工程与应用 ›› 2010, Vol. 46 ›› Issue (30): 148-152.DOI: 10.3778/j.issn.1002-8331.2010.30.044

• 图形、图像、模式识别 • 上一篇    下一篇

噪音人脸图像的总间隔v最小类内方差SVM分类

杨 冰,王士同   

  1. 江南大学 信息工程学院,江苏 无锡 214122
  • 收稿日期:2009-12-16 修回日期:2010-02-08 出版日期:2010-10-21 发布日期:2010-10-21
  • 通讯作者: 杨 冰

Total margin v minimum class variance SVM for noisy face classification

YANG Bing,WANG Shi-tong   

  1. School of Information Technology,Jiangnan University,Wuxi,Jiangsu 214122,China
  • Received:2009-12-16 Revised:2010-02-08 Online:2010-10-21 Published:2010-10-21
  • Contact: YANG Bing

摘要: 提出总间隔v最小类内方差支持向量机(TM-v-MCVSVMs),用于解决含有噪音人脸图像的分类问题,它综合了最小类内方差支持向量机(MCVSVMs)和总间隔v-支持向量机(TM-v-SVM)的优点。给出了TM-v-MCVSVMs在小样本问题和非线性分类问题中的解决方法。经初步的实验验证,在含有噪音人脸图像的分类问题中,TM-v-MCVSVMs获得了比MCVSVMs和TM-v-SVM更好的分类性能。

关键词: 支持向量机(SVM), 最小类内方差支持向量机(MCVSVMs), 总间隔v-支持向量机(TM-v-SVM), 人脸识别, 主成分分析(PCA), 核主成分分析(KPCA)

Abstract: In this paper,the Total Margin[v]Minimum Class Variance Support Vector Machines(TM-[v]-MCVSVMs) as the improved version of Minimum Class Variance Support Vector Machines(MCVSVMs) is presented for noisy face recognition,which integrates the advantages of MCVSVMs with TM-[v]-SVM.The discussions about the proposed TM-[v]-MCVSVMs for “small sample size” problem and nonlinear classifications are also given.The experimental results about noisy face classification demonstrate that the proposed TM-v-MCVSVMs has better classification performance than both MCVSVMs and TM-v-SVM.

Key words: Support Vector Machine(SVM), Minimum Class Variance Support Vector Machines(MCVSVMs), Total Margin vsupport Vector Machine(TM-v-SVM), face recognition, Principal Component Analysis(PCA), Kernel Principal Component Analysis(KPCA)

中图分类号: