基于分类稀疏表示的人脸表情识别

为挖掘不同人脸表情图像的统计特性差异，提出一种基于分类稀疏表示的表情识别算法。首先通过对不同类别表情图像的字典学习，构建满足各类表情图像统计特性的基函数子集，进而采用Lasso算法获得表情图像在由基函数集所张成特征子空间中的稀疏表示，最后通过比较表情图像在各基函数子集上的重构误差实现不同表情的分类识别。基于JAFFE人脸表情数据库的实验结果表明，该算法可以有效克服人脸身份对表情识别的影响，具有较高的表情识别率和鲁棒性。

Face Expression Recognition Based on Class-depend Sparse Representation

In this paper, a kind of face expression recognition method based on class-depend sparse representation is proposed to find the statistic property difference between the different face expression images. Firstly, the base function sets corresponding to each class of expression are constructed by dictionary learning based on class-depend images, and the sparse representation of test image on different base fimction set can be acquired by Lasso algorithm. Finally, the reconstructive errors between test image and its sparse representation in each class base function set are used to determine its classification. The test experimental result on JAFFE face database shows that this method is effective to anti-interference from person pattern, and has a high recognition accuracy and robust for face expression.