基于Mahalanobis距离进行人脸表情的识别  

Mahalanobis distance based facial expression recognition

作　　者：黄鹤鸣[1] 赵晨星[2] 

机构地区：[1]青海师范大学物理系,青海西宁810008 [2]青海藏文信息技术研究所,青海西宁810008

出　　处：《青海师范大学学报（自然科学版）》2005年第4期46-49,共4页Journal of Qinghai Normal University(Natural Science Edition)

基　　金：信息产业部;信部运[2002]393号

摘　　要：提出了利用Mahalanobis距离进行人脸表情识别的方法.首先将待分类的图像样本集进行坐标变换,使得变换以后类间离散度尽可能大而类内离散度尽可能小,即使变换以后的Fisher准则函数取得极大值,在新的坐标下求每个待分类样本到各类均值向量的Mahalanobis距离,从而将待分类的样本归到Mahalanobis距离最小的类中去,通过实验得到了平均80.25%的识别率.In this paper, we proposed Mahalanobis distance based technique to classify four facial expressions namely： neutral, joy, surprise and sad. First we＇ ve mapped all the sample vectors to a new coordinate by using the optimal linear transform matrix W that makes Fisher criterion function maximum. Then in this new coordinate we calculate the Mahalanobis distance of every sample vector to the mean vectors of four classes and classify it to the class to which its Mahalanobis distance is the least. And the average recognition rate we＇ve got is 80.25%.

关 键 词：人脸表情识别 MAHALANOBIS距离 类内离散度距阵 类间离散度距阵 FISHER准则 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]