遮挡条件下的鲁棒表情识别方法

提出了一种对面部遮挡具有鲁棒性的表情识别方法.首先,基于鲁棒主成分分析(RPCA,Robust Principal Component Analysis)对待识别人脸进行重构,并对重构人脸和待识别人脸的差值图像进行显著性检测得到面部遮挡区域;其次,将待识别人脸的遮挡区域由RPCA重构人脸的相应区域进行替换,并由权值更新的AdaBoost分类器对遮挡区域重构后的人脸进行表情识别.在BHU(Beihang University)人脸表情数据库和日本女性表情数据库上进行了各种遮挡情况下的表情识别实验,获得了比AdaBoost方法更好的识别结果,说明基于RPCA和AdaBoost的表情识别方法对多种面部遮挡具有较强的鲁棒性.

Robust facial expression recognition under occlusion condition

A novel method for facial expression recognition which is robust to facial occlusion was proposed. Firstly, the face to be recognized is reconstructed using robust principal component analysis(RPCA), and saliency detection was used on the difference image of reconstructed face and the face to be recognized to obtain the facial occlusion region. Secondly, facial occlusion region of the face to be recognized was reconstructed by the reconstructed face using RPCA, and a novel reweighted AdaBoost classifier was used on the face after occlusion region reconstruction for facial expression recognition. Finally, facial expression recognition experiments were implemented in different occlusion conditions on Beihang university(BHU) facial expression database and Japanese female facial expression database and gained better recognition results than Adaboost method, showing that this method based on RPCA and AdaBoost is robust to kinds of facial occlusions.