摘要
提出一种基于稀疏保留投影的人脸表情识别方法,称之为SPP。与局部保留投影(LPP)不同,SPP通过稀疏重构处理,在保留表情稀疏重构信息的同时也保留表情局部邻信息,这样可从原始表情数据中提取更多、更有效、更具判决性的内在表情特征,获得的投影也更稳定。基于CED-WYU(1.0)和JAFFE两个表情数据库的识别结果表明,基于SPP的特征提取方法能有效地提高识别率。
A facial expression recognition method based on Sparse Preserving Projection（SPP） was proposed in this paper.Unlike Local Preserving Projection（LPP）,SPP aims to preserve the sparse reconstructive relationship of the expression data,simultaneously,local neighborhood information was preserved during the dimensionality reduction procedure.SPP could extract more useful and discriminatal expression features from the orginal expression data,and the obtained projections were more stable.Experimental result on CED-WYU（1.0）and JAFFE show that SPP is an effective method for improving the recognition accuracy.
出处
《计算机应用》
CSCD
北大核心
2010年第12期100-101,125,共3页
journal of Computer Applications
关键词
局部保留投影
稀疏保留投影
表情识别
Local Preserving Projection（LPP）
Sparse Preserving Projection（SPP）
expression recognition
