基于融合DCT和LBP特征提取的面部表情识别

Facial Expression Recognition Based on Fusion Feature Extraction of DCT and LBP

摘要:

针对离散余弦变换(DCT)只能提取面部表情图像的全局特征, 而忽略了临近像素之间的关系、不能提取纹理特征信息、不能准确区分相似表情等问题, 提出一种融合离散余弦变换方法和局部二值模式(LBP)特征的表情特征提取方法。该方法首先将人脸图像经过DCT获得的低频系数作为表情的全局特征; 然后用LBP对贡献率较大的嘴部、眼睛区域进行局部纹理特征提取, 通过将LBP提取到的局部纹理特征与DCT提取到的全局特征进行融合, 从而得到更有效的表情特征; 最后利用支持向量机(SVM)进行识别。实验结果表明: 该方法比单独使用DCT方法提取的表情特征更有利于识别, 提高了表情识别的准确性, 并将这个表情识别方法用于智能轮椅的控制上, 收到了良好的效果。

Abstract:

To the problem that the discrete cosine transform (DCT) only can extract the global characteristics of facial expression while ignore the relationship between the adjacent pixels, so the texture feature information can not be extracted and similar expression can not be accurately distinguished. A novel method by fusing discrete cosine transform (DCT) and local binary pattern (LBP) features was proposed for expression recognition in this research. The primary information of the face image was centralized in a small number of DCT coefficients, which were used as the global feature of the expression. Then LBP was used to extract the local features of the mouth and eyes area, which contributes most to facial expression recognition. Fusing the global and local texture feature would be more effective for facial expression recognition. Finally, support vector machine (SVM) was used to make feature fusion so as to complete facial expression recognition. Experimental results indicate that this method is better than individual DCT, which improves the accuracy of expression recognition, and can be used in the control of intelligent wheelchair successfully.