人脸样貌独立判别的协作表情识别算法

发布时间：2018-04-18 17:51

本文选题：表情识别 + 稀疏表示　；参考：《计算机应用研究》2017年09期

【摘要】：为了降低样貌、姿态、眼镜以及表情定义不统一等因素对人脸表情识别的影响,提出一种人脸样貌独立判别的协作表情识别算法。首先,采用自动的人脸检测算法定位、对齐视频每帧的人脸区域,并从人脸视频序列中选择峰值表情的人脸;然后,采用峰值人脸与某个表情类内的所有人脸产生表情类内差异人脸信息,并通过计算峰值表情人脸与表情类内差异人脸的差异信息获得协作的表情表示;最后,采用基于稀疏的分类器与表情表示决定每个人脸表情的标签。采用欧美与亚洲人脸的数据库进行仿真实验,结果表明本算法获得了较好的表情识别准确率,对不同样貌、佩戴眼镜的人脸样本也具有较好的识别效果。
[Abstract]:In order to reduce the influence of facial features, posture, glasses and facial expression definitions on facial expression recognition, a collaborative facial expression recognition algorithm based on independent recognition of facial features is proposed.First, the automatic face detection algorithm is used to locate the face region of each frame of the video, and the human face with the peak facial expression is selected from the video sequence.Using the peak face and all the faces in a certain facial expression class to produce intra-facial expression difference information, and by calculating the difference information between the peak facial expression face and the facial expression class difference information, the cooperative expression of facial expression is obtained. Finally,A sparse classifier and facial expression are used to represent the labels that determine each person's facial expression.By using the database of European, American and Asian faces, the simulation results show that the algorithm has a good accuracy of facial expression recognition, and it also has a good recognition effect for face samples with different looks and glasses.
【作者单位】：枣庄学院信息科学与工程学院;中国石油大学(华东)理学院;
【基金】：山东省高校科研计划研究项目(J15LN81,J13LN56) 枣庄学院大学生研究训练计划项目(2015061) 山东省教育厅项目(J13LN56)
【分类号】：TP391.41

【相似文献】