判别性完全局部二值模式人脸表情识别  

Facial expression recognition based on discriminative CLBP

作　　者：周宇旋 吴秦[1] 梁久祯[1] 王念兵[1] 李文静[1] ZHOU Yuxuan;WU Qin;LIANG Jiuzhen;WANG Nianbing;LI Wenjing(Institute of Intelligent Systems and Network Computing, School of Internet of Things Engineering, Jiangnan University,Wuxi, Jiangsu 214122, China)

机构地区：[1]江南大学物联网工程学院智能系统与网络计算研究所,江苏无锡214122

出　　处：《计算机工程与应用》2017年第4期163-169,194,共8页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61202312;No.61170121);教育部留学回国人员科研启动基金项目

摘　　要：针对完全局部二值模式(CLBP)存在直方图维数过高和特征冗余,会导致识别速度降低和识别率低的问题,提出基于有判别力的完全局部二值模式(Discriminative completed LBP,dis CLBP)的人脸表情识别算法。首先,对人脸表情图像进行预处理获得表情子区域;然后提取表情子区域和整幅图像的dis CLBP特征,针对不同的表情筛选出不同的表情特征,再将筛选出的表情子区域特征直方图融合;最后用最近邻分类器进行分类识别。该算法在CK人脸表情库上进行实验的平均识别率为97.3%。To overcome high dimension and characteristic redundancy of CLBP features which result in the decrease of recognition speed and low recognition rate, this paper proposes an algorithm for facial expression recognition based on discriminative Completed Local Binary Pattern(disCLBP). First of all, the facial expression sub-regions are isolated by a preprocessing step; then sub-regions and the global facial expression images are extracted by disCLBP, and different features are filtered for different expressions. After that cascade histograms are generated by associating the histograms of the filtered features of the images and the sub-regions. In the end, filtered features are classified by nearest neighbor classifier. Theaccuracy of the experiment on the facial expression database of CK is 97%.

关 键 词：完全局部二值模式(CLBP) 有判别力的完全局部二值模式(disCLBP) 表情识别 最近邻 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]