一种基于Fisher准则的二维主元分析表情识别方法  

A Novel Two-Dimensional PCA Facial Expression Recognition Method Based on Fisher Ratio

作　　者：程剑[1] 应自炉[1] 

机构地区：[1]五邑大学信息学院,广东江门529020

出　　处：《五邑大学学报（自然科学版）》2006年第2期42-46,共5页Journal of Wuyi University(Natural Science Edition)

基　　金：广东省自然科学基金项目(NO032356);北京大学视觉与听觉信息处理国家重点实验室开放课题

摘　　要：提出了一种基于Fisher准则进行特征选择的二维主元分析表情识别方法.首先对训练样本做二维主元分析,然后再根据Fisher准则,按Fisher比的大小选择特征向量作为投影轴,最后用最近邻方法进行分类.在JAFFE人脸表情静态图像库上进行实验,与按特征值的大小来选择特征向量相比,该方法更有效.In this paper, a novel two-dimensional principal component analysis （2-D PCA） method based on Fisher ratio is proposed for facial expression recognition. First, 2-D PCA is applied to training sample expression images. Then, the feature vectors are chosen according to their Fisher Ratio and are used as projection vectors, Finally, NN is used as the classifier to classify facial expressions. The results of experiments on JAFFE facial expression database show that the method can improve the recognition rate.

关 键 词：FISHER准则 二维主元分析 表情识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]