基于D-S证据理论的表情识别技术  

Research on Facial Expression Recognition Technology Based on D-S Ttheory

作　　者：王嵘[1] 马希荣[1] 

机构地区：[1]天津师范大学计算机与信息工程学院,天津300387

出　　处：《计算机科学》2009年第1期231-233,共3页Computer Science

基　　金：天津市自然科学基金项目(编号:06YFJMJC00100);天津师范大学青年基金(52LJ88)资助

摘　　要：在情感计算理论基础上,提出了基于D-S理论的信息融合的表情识别技术,设计并实现了系统IFFER。在表情识别模块中的分类器训练采用JAFFE表情库。识别中首先利用色度匹配及亮度匹配将人脸图像进行眼部及嘴部的分割,再分别用训练好的眼部SVM分类器及嘴部SVM分类器进行识别,将识别后的结果利用D-S证据理论进行融合。实验结果表明,对分割后的两部分图像进行识别,无论从训练上还是识别上,数据的维数都大大减少,提高了效率。在识别率上,融合后的结果相对于融合前的有显著的提高。This paper proposed a kind of facial expression recognition technology based on the D-S theory of information fusion, in the background of affective computing theory, moreover designed and implemented the system IFFER. In the part of facial expression recognition, the training for the classification uses JAFFE expression library. Using color match and brightness match, this method divides face image into the eye and mouth images, then identifies facial expression respectively by trained eye SVM classifier and mouth SVM classifier, in the end, fuses the identification using D-S theory. The experimental results show that recognizing the two parts of image,in terms of training or recognition, the data dimension has greatly reduced and increased efficiency. In recognition rate, the fusion of the results has more improved than the pre-fusion.

关 键 词：情感计算 信息融合 表情识别 D-S理论 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置] TP391.4[自动化与计算机技术—控制科学与工程]