摘要
提出了一种运用SRC(Sparse Representation based Classification)在个体子空间里,进行表情识别的新方法。用Gabor滤波器,提取表情图像的特征。进行稀疏分解,得到稀疏表示系数。根据稀疏系数确定待测图像所在的子空间,在子空间里,完成表情识别。这种方法较好地避免了不同个体对表情识别的干扰,从而提高了表情识别的正确率。在Cohn-Kanade和JAFFE人脸库上的表情识别实验表明,该方法对表情识别非常有效。
A novel method to recognize facial expression in individual subspace using SRC(Sparse Representation based Classification)is proposed. Gabor filter is used to extract features of test expression image. The sparse representation coefficients of test image are gained by sparse decomposition. The individual subspace where test image lies in is found according to its sparse representation coefficients and its expression category is recognized in the subspace. The method eliminates preferably the disturbance of object without same identity to expression recognition. So recognition rates of expressions have been improved effectively. Experiments on the Cohn-Kanade face database and JAFFE face database show the proposed method achieves high performance to expression recognition.
出处
《计算机工程与应用》
CSCD
2014年第12期33-37,共5页
Computer Engineering and Applications
基金
湖南省自然科学基金(No.12JJ3061)
湖南省优秀青年基金(No.10B074)
关键词
表情识别
稀疏表示
GABOR滤波
子空间
expression recognition
sparse representation
Gabor filtering
subspace