提出了一种基于图像优化局部保留投影的人脸表情识别方法.该算法在降维过程中将图像结构信息融入到LPP目标函数中,称之为GOLPP.与LPP不同,GOLPP通过降维处理,获得图像结构信息的同时将投影最优化,这样可从原始表情数据中提取更具判决性的表情信息.基于JAFFE和CED-WYU (1.0)两个表情数据库的识别结果表明,基于GOLPP的特征提取方法能有效地提高识别率.%A facial expression recognition method based on Graph-Optimized Locality Preserving Projections(GOLPP) is proposed in this paper.Unlike LPP,GOLPP incorporates graph construction into the LPP objective function in dimension reduction process, thus obtains a simultaneous learning framework for graph construction and projection optimization, GOLPP can extract more useful and discriminatal expression features from the original expression data.Experimental result on JAFFE and CED-WYU show that GOLPP is an effective method for improving the recognition accuracy.
展开▼