Abstract: Discriminant analysis is one of crucial issues for the statistic-based face recognition method.This paper proposes a novel large margin nonlinear discriminant analysis of minimization within class scatter.The underlying idea is that the kernel trick is used firstly to project the original samples into an implicit space called feature space by nonlinear kernel mapping，the kernel trick is used to improve the traditional large margin classifier algorithm，moreover，the theory of reproducing kernel in the new feature space is used to obtain the optimal kernel discriminant vectors with which the kernel within scatter is kept as small as possible.Finally，the proposed method is tested on ORL face database，the results prove this arithmetic optimal.

Key words: large margin classifier, Support Vector Machine（SVM）, nonlinear discriminant analysis, kernel function, within class scatter

摘要： 提出了一种新的非线性鉴别分析算法——极小化类内散布的大间距非线性鉴别分析。该算法的主要思想是将原始样本映射到更高维的空间中，利用核技术对传统的大间距分类算法进行改进，在新的高维空间中利用再生核技术寻找核鉴别矢量，使得在这个新的空间中核类内散度尽可能的小。在ORL人脸数据库上进行实验，分析了识别率及识别时间，结果表明该方法具有一定优势。

关键词: 大间距分类器, 支持向量机, 非线性鉴别分析, 核方法, 类内散布