在正交化Fisher鉴别分析的基础上提出了正交化核Fisher鉴别分析方法。该方法具有理论简单、计算方便、特征表示能力强等优点。在CENPARMI手写体阿拉伯数字数据库、NUST603HW手写体汉字样本库和FERET人脸图像数据库上的仿真实验结果表明,正变化Fisher鉴别和正交化核Fisher鉴别方法在特征抽取能力和特征抽取效率方面均分别优于FoleySammon鉴别和核Foley-Sammon鉴别方法。
Based on the orthogonalized Fisher discriminant, the orthogonalized kernel Fisher discriminant method is proposed. The method is simple in theory and easy to be calculated, and has a good feature representation. Experimental results conducted on CENPARMI hand written Arabic digit database, NUST603HW hand written Chinese character database, and FERET face image database show that the orthogonalized Fisher discriminant and the orthogonalized kernel Fisher discriminant are better than the Foley-Sammon discriminant and the kernel Foley-Sammon discriminant in the recognition efficiency.