基于多姿态多状态面部情绪模型的表情识别  

A method for the recognition of a variety of facial expression

作　　者：陈国社[1] 张青[2] 李凡[1] 

机构地区：[1]华中科技大学计算机科学与技术学院,湖北武汉430074 [2]黄冈师范学院物理与电子技术系,湖北黄冈436100

出　　处：《华中科技大学学报（自然科学版）》2004年第8期60-62,共3页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家高性能计算基金资助项目 (0 0 30 3)

摘　　要：介绍了面部表情识别的主要步骤 ,在此基础上提出了一种多姿态多状态的面部情绪模型 .该系统利用图像质量的评价结果来决定面部检测的方法 ,通过中性脸的检测来实现表情边界的分割 ,从而把提取的表情运动特征参数作为BP神经网络的输入 ,进而实现面部表情的识别 .实验表明 ,该系统具有很好的识别效果 .The primary steps of facial expression recognition were introduced; the multi pose and multi state facial emotion models were proposed. The method of face detection by using the result of image quality estimation was selected. Segmentation of expression border was realized by the detection of neural face. By taking the feature parameters of the extracted expression motion as the input of BP neural network, the facial expression was recognized. The results showed that the system was characterized by good recognition.

关 键 词：面部表情识别 面部检测 表情特征区域 多姿态 多状态 面部情绪模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]