本文已被：浏览 1655次 下载 804次
投稿时间：2014-03-27
投稿时间：2014-03-27
中文摘要: 提出了一种基于三维特征点模型匹配的人脸表情识别算法.首先,以混合树结构建模不同姿态脸部特征组合,实现特征点定位,并建立脸部三维几何特征模型;然后,基于广义普鲁克分析原理对特征点模型进行归一化计算,并建立具有鲁棒性的匹配特征;最后,采用支持向量机实现表情的训练和识别.通过对BU-3DFE数据库中6种脸部表情的测试,表明该算法能够达到很好的识别效果.
Abstract:A facial expression recognition method based on 3D landmark model matching is proposed. Firstly, facial landmarks are extracted based on a mixtures-of-trees structured model, and then a facial landmark model is built. Then Generalized Procrustes Analysis is used to do the model alignment in order to get the robust matching features. Finally, facial expression recognition is accomplished by SVM classification. Experiments are conducted on BU-3DFE dataset with six different facial expressions. The result shows that the algorithm has good performance on facial expression recognition.
keywords: 3D landmark matching facial expression recognition
文章编号：20140412 中图分类号： 文献标志码：
基金项目:国家自然科学基金(61302151);上海市自然科学基金(13ZR1455100)
引用文本：
邵洁,董楠.基于三维特征点模型的人脸表情识别算法[J].上海电力大学学报,2014,30(4):352-356,364.
SHAO Jie,DONG Nan.Facial Expression Recognition Based on 3D Landmark Model[J].Journal of Shanghai University of Electric Power,2014,30(4):352-356,364.
邵洁,董楠.基于三维特征点模型的人脸表情识别算法[J].上海电力大学学报,2014,30(4):352-356,364.
SHAO Jie,DONG Nan.Facial Expression Recognition Based on 3D Landmark Model[J].Journal of Shanghai University of Electric Power,2014,30(4):352-356,364.