摘要： 提出一种基于稀疏保留投影的人脸表情识别方法。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，可从原始表情数据中提取更多有效且具判决性的内在表情特征，获得的投影也较稳定。基于JAFFE和CED-WYU 2个表情数据库的识别结果表明，该方法能有效提高人脸表情识别率。

关键词: 数据降维, 局部保留投影, 稀疏保留投影, 人脸表情识别

Abstract: A facial expression recognition method based on Sparse Preserving Projections(SPP) is proposed in this paper. It aims to preserve the sparse reconstructive relationship of the expression data, simultaneously, local neighborhood information is preserved during the dimension reduction procedure. It can extract more useful and discriminatal expression features from the orginal expression data, and the obtained projections are more stable. Experimental result on JAFFE and CED-WYU show that this method is an effective method for improving the recognition accuracy

Key words: dimension reduction, Locality Preserving Projections(LPP), Sparse Preserving Projections(SPP), facial expression recognition

中图分类号: 

• TP391