一种加权的核Fisher鉴别分析在人脸识别中的应用  

The Application of a Weighted Kernel Fisher Discriminant Analysis Applied in Face Recognition

作　　者：刘晓亮[1] 王福龙[1] 黄诚[1] 曾爱华 

机构地区：[1]广东工业大学应用数学学院,广东广州510006

出　　处：《广东工业大学学报》2009年第4期65-69,共5页Journal of Guangdong University of Technology

摘　　要：在非线性空间中采用加权的最大散度差鉴别准则函数,该方法不仅有效地抽取了人脸图像的非线性特征,而且在特征空间H中,使用权函数重新构造了类间散度矩阵和类内散度矩阵,从而优化了核的最大散度差准则函数。最后在ORL和Yale人脸数据库上的实验结果验证了本文方法的有效性。A weighted kernel maximum scatter difference discriminating criterion is developed for extraction of nonlinear features. The proposed method can be used to extract nonlinear features for faces effectively anf to reconstruct between-class scatter matrix and within-class scatter matrix by weighted schemes. Therefore, it can modify the kernel maximum scatter difference discriminating criterions. Experiments performed on ORL and Yale face database verify the effectiveness of the proposed method.

关 键 词：人脸识别 Fisher非线性鉴别分析 最大散度差鉴别准则 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]