Computer Engineering and Applications ›› 2011, Vol. 47 ›› Issue (23): 208-211.

• 图形、图像、模式识别 • Previous Articles     Next Articles

Supervised discriminant analysis derived from dissimilarity

YU Zhenzhou,WANG Zhengqun,CHEN Guanghua   

  1. School of Information Engineering,Yangzhou University,Yangzhou,Jiangsu 225009,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-08-11 Published:2011-08-11



  1. 扬州大学 信息工程学院,江苏 扬州 225009

Abstract: This paper proposes a novel supervised discriminant analysis method derived from dissimilarity.Combining pattern local and global information,the new within-class and between-class scatter weight matrixes are defined.They represent dissimilarity of within-class sample and between-class sample respectively.The new within-class scatter and between-class scatter matrix are derived from the new scatter weight matrixes.The optimal transformation matrix is determined according to the Fisher criterion.The experimental results on YALE and AR face image database show that the proposed method outperforms the traditional approaches.

Key words: scatter weight matrix, Fisher criterion, dissimilarity, global and local information, face recognition

摘要: 提出了相异度导引的有监督鉴别分析方法(D-SDA)。结合模式局部信息和全局信息,定义了类内散度权重矩阵[RW]和类间散度权重矩阵[RB],分别表示类内样本的相异度、类间样本的相异度。由[RW]、[RB]导出类内散度矩阵[SW]和类间散度矩阵[SB],根据Fisher鉴别准则函数确定最优变换矩阵。在YALE和AR人脸图像库上的实验验证了这一算法的有效性。

关键词: 散度权重矩阵, Fisher判别准则, 相异度, 全局与局部信息, 人脸识别