一种人脸图象自动识别方法及实验

A Novel Method on Face Recognition

为了很好解决小样本情况下线性鉴别矢量的求解问题,定义了Fisher极小准则函数与广义Fisher极小准则函数,并从理论上说明了二者的等价性。基于广义Fisher极小准则函数,推导出一个新的广义特征方程。具有最多鉴别信息的鉴别矢量可通过求解新方程得到。人脸图象的识别结果表明,本文方法具有较优性能。

Fisher minimum criterion and generalized Fisher minimum criterion are defined. In fact, the two criteria are equivalent. Based on the generalized Fisher minimal criterion, a novel eigen equation is proposed. Linear discriminant vectors with most discriminant information can be obtained through the novel eigen equation. Experiment on a face database shows that the method used in this paper is excellent.