融合人脸局部区域的表情识别  

Facial Expression Recognition By Fusing Local Facial Regions

作　　者：应自炉[1,2] 张有为[1,2] 李景文[1] 

机构地区：[1]北京航空航天大学电子信息工程学院,北京100083 [2]五邑大学信息学院,广东江门529020

出　　处：《信号处理》2009年第6期963-966,共4页Journal of Signal Processing

摘　　要：本文提出了一种融合人脸各个局部区域的表情识别方法,首先提取眼睛、鼻子和嘴巴三个局部区域,并对这三个局部区域和整个人脸区域分别进行表情识别,得到四个独立的识别结果,再利用融合策略对四个结果进行决策融合,所提方法在日本女性表情数据库JAFFE库和五邑大学人脸表情数据库CED_WYU库上进行了广泛的实验,结果表明,通过融合人脸各局部区域的识别结果,表情识别率可以得到显著提高。A novel method for facial expression recognition by fusing local facial regions is proposed. The new approach first extracts local regions such as mouth, nose and eyes. Then expression recognition is carried out based on these regions as well as on the whole face. The final facial expression recognition result is obtained by fusing recognition results based on local facial regions. The pro- posed approach is experimented extensively on both Japanese female facial expression database JAFFE and Chinese Expression Database Wuyi university facial expression database CED_WYU. The experiment results show that the facial expression recognition rates increase obviously compared with those based only on local regions or only the whole face and prove the effectiveness of the proposed fusion approach.

关 键 词：人脸表情识别 融合 DS理论 决策模板 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]