基于稀疏表示的低分辨率人脸疲劳表情识别  

Method of Low Resolution Facial Fatigue Expression Recognition Based on Sparse Representation

作　　者：张灵[1] 田小路[1] 罗源[1] 常捷[1] 吴勇[1] 

机构地区：[1]广东工业大学计算机学院,广州510006

出　　处：《计算机科学》2016年第9期305-309,共5页Computer Science

基　　金：广东省自然科学基金(2014A030310169);广州市科技计划(2014Y2-00211)资助

摘　　要：为了有效提高低分辨率图像的人脸疲劳表情识别性能,提出一种基于稀疏表示的低分辨率人脸疲劳表情的识别方法。首先,采用肯德尔和谐系数可信度分析法构建了低分辨率人脸疲劳表情图像库TIREDFACE。其次,通过图像库中的低分辨率样本疲劳表情图像进行稀疏表示,再利用压缩感知理论寻求低分辨率测试样本的最稀疏解,采用求得的最稀疏解实现低分辨率人脸疲劳表情的分类。在低分辨率人脸视觉特征的疲劳表情图像库TIREDFACE的实验测试结果表明,将该方法用于低分辨人脸疲劳表情识别,性能优于线性法、最近邻法、支持向量机以及最近邻子空间法。可见,该方法用于低分辨率人脸疲劳表情识别时识别效果较好,精确度较高。In order to effectively improve the performance of facial fatigue expression recognition on the low resolution image, a method of fatigue facial expression recognition based on sparse representation was proposed. Firstly, the relia- bility analysis method of Kendall coefficient of concordance is used to construct the low-resolution facial fatigue expres- sion database TIREDFACE. Secondly, the sparse representation of the low resolution facial fatigue expression images of the identified test samples in the database is sought, and then the compressed sensing theory is used to seek their spar- sest solution. Finally, according to the sparsest solution, the low-resolution facial fatigue expression classification is per- formed. Experimental results on TIREDFACE database show that the low resolution facial fatigue expression perfor- mance obtained by this method is much better than the linear classifier, the nearest neighbor （NN）, support vector ma- chine （SVM） and the nearest subspace （NS）. Therefore, the proposed method on the low resolution facial fatigue ex- pression recognition tasks achieves better performance and high accuracy.

关 键 词：稀疏表示 压缩感知 疲劳表情 基于稀疏表示分类 肯德尔和谐系数 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]