摘要
提出一种基于优化局部保留投影(OLPP)的人脸表情识别方法。OLPP方法在降维过程中将图像结构信息融入LPP目标函数,通过降维处理,在获得图像结构信息的同时将投影最优化,从而能从原始表情数据中提取更多更具判决性的有效表情信息。JAFFE和CED-WYU(1.0)2个表情数据库的识别结果表明,基于OLPP的特征提取方法能有效提高识别率。
A facial expression recognition method based on Optimized Locality Preserving Projections(OLPP) is proposed. Unlike Locality Preserving Projections(LPP), OLPP incorporates graph construction into the LPP objective function in dimension reduction process, thus obtains a simultaneous learning framework for graph construction and projection optimization. OLPP can extract more useful and discriminating expression features from the original expression data. Experimental result on JAFFE and CED-WYU(1.0) shows that OLPP is an effective method for improving the recognition accuracy.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第4期210-211,220,共3页
Computer Engineering
关键词
局部保留投影
优化局部保留投影
表情识别
Locality Preserving Projections(LPP)
Optimized Locality Preserving Projections(OLPP)
expression recognition