基于改进LBP的人脸表情识别  

Facial Expression Recognition Based on Improved LBP

作　　者：姜锐[1] 许建龙[1] 张爱朋[1] 

机构地区：[1]浙江理工大学信息学院,杭州310018

出　　处：《浙江理工大学学报（自然科学版）》2013年第4期546-549,共4页Journal of Zhejiang Sci-Tech University(Natural Sciences)

摘　　要：局部二元模式即LBP(local binary patterns),是一种有效的纹理描述算子,能够很好地提取人脸表情特征信息。针对原始LBP算子仅考虑中心像素点与邻域像素点的灰度差异的问题,对其进行了改进,提出了多重中心化二值模式MLBP(multiple local binary patterns),并将改进后的MLBP算子进行人脸表情识别,通过改进前后在JAFFE人脸库的实验比较,该方法在识别率上取得了较好的效果。LBP（Local Binary Pattern） is an effective texture description operator whmh can well extract facial expression feature information. This paper improves the original LBP operator in allusion to the problem that it only considers gray difference between central pixel and neighborhood pixel, puts forward multiple local binary patterns MLBP and applied the improved MLBP operator in facial expression recognition. Through the experimental comparison in JAFFE face database before and after improvement, this method achieves a good effect in terms of the recognition rate.

关 键 词：面部表情识别 局部二值模式 特征提取 支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]