基于零空间核决策分析的人脸识别研究

Research of face recognition based on null space kernel discriminant analysis

针对零空间线性决策分析方法难以揭示人脸图像空间中数据的非线性结构的问题,提出了一种零空间核决策分析方法,详细介绍了该方法的推导过程及求解步骤。测试结果表明,该方法能够在核空间中提取类内离散度矩阵的零空间,并且最大程度上去除类间离散度矩阵的零空间,新提取的特征能够有效地用来进行人脸识别。

In view of problem that method of null space linear discriminant analysis is difficult to reveal nonlinear structure of data of face image space, the paper proposed a method of null space kernel discriminant analysis. It introduced derivation process and solving steps of the method in details. Test result shows that the method can extract null space of scatter matrix within class in kernel space and remove null space of scatter matrix inter classes to the utmost extent, the new extracted feature can be used to make face recognition effectively.