基于面部结构的表情识别  

Component-based facial expression recognition

作　　者：李淑婧[1] 嵇朋朋 邓健康[1] 孙玉宝[1] 刘青山[1] 

机构地区：[1]南京信息工程大学信息与控制学院,南京210044

出　　处：《计算机应用研究》2015年第3期917-921,941,共6页Application Research of Computers

基　　金：国家自然科学基金资助项目(61272223;61300163);江苏省杰出青年基金资助项目(BK2012045);江苏省自然科学基金资助项目(BK20131003);模式识别国家重点实验室开放课题基金资助项目(201204234)

摘　　要：心理学上的研究表明,面部表情变化主要集中在眼睛、眉毛、鼻子、嘴巴等面部器官上。受其启发,提出一种基于面部结构的表情识别方法,重点分析眼睛、眉毛、鼻子、嘴巴等关键区域的联动变化来分析表情。首先,使用鲁棒的判别响应图拟合(discriminative response map fitting,DRMF)方法自动检测出对识别人脸表情最为关键的局部人脸区域,即眼睛、眉毛、鼻子和嘴巴的部分;然后从这些关键部分中提取Haar特征;最后采用Boosting学习和联动机制,学习得到基于联合Haar特征的表情分类器。在CMU表情数据库和JAFFE表情数据库上的实验结果表明了上述方法的良好性能,即基于面部构件识别表情的方法获得了与手工精准标注人脸面部构件识别表情方法相近的识别效果。Psychological research shows that facial expressions are mainly focused on the eyes, eyebrows, nose, mouth and other facial organs. So this paper proposed a component-based facial expression recognition method, which focused on analyzing the linkage variations among eyes, eyebrows, nose, mouth, etc, First, it used the robust discriminative response map fitting meth- od to automatically detect eyes, eyebrows, nose and mouth. Then it extracted Haar-like features from these key facial compo- nents. Finally it used Boosting learning with a linkage mechanism to get the joint Haar feature based expression classifier. The experimental results on CMU database and JAFFE database show the good performance of the proposed approach,namely com- ponent-based facial expression recognition obtains similar recognition accuracy compared with the facial expression recognition method of marking facial critical areas manually.

关 键 词：人脸表情识别 判别响应图拟合 联合Haar-like特征 Boosting学习 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]