中文摘要：

为充分利用视频中人脸表情与中性表情的差异,提出了一种新的对非特定人脸的表情识别方法.该方法针对低复杂度的视频表情识别应用场景,利用参考中性表情的特征点偏移角表征被测表情的变化信息,同时利用二维主成分分析（2DPCA）法提取被测表情帧的二维主成分特征,从而综合使用表情的动态和静态特征,并使用支持向量机分类器进行表情分类识别.在JAFFE人脸表情库上的实验结果表明,相对于仅使用2DPCA的静态图像表情识别方法,文中所提方法的人脸表情识别准确率平均提高7％.

英文摘要：

In order to fully explore the difference in facial expression in videos, a novel facial expression recogni- tion approach for non-repeatable face, which utilizes both dynamic and static features to recognize the facial expres- sion in videos in low-complexity scenarios, is proposed. In this approach, the feature point offset angle of neutral expression is presented as a feature to represent the difference information of measured facial expression, and simul- taneously, the two-dimension principal component features are extracted via two-dimension principal component analysis （2DPCA）. Then, the combined dynamic-static information is sent to support vector machine （SVM） clas- sifiers to implement the facial expression recognition. Experimental results on JAFFE database indicate that, in comparison with the existing method which only uses the static features extracted via 2DPCA, the proposed method may result in 7 % of increment in facial expression recognition accuracy.