一种基于Radon变换的微表情识别算法

A Micro-expression Recognition Algorithm Based on Radon Transform

微表情是人们处在一些与平时生活环境不同的高强度环境下试图控制和掩饰的情感表现,也是一种不曾意识到的瞬时脸部表情,持续时间短,强度弱。为了提高其准确率,提出了基于Radon变换的微表情识别算法。首先,对数据库中的视频序列进行灰度归一化、尺寸归一化和二维主成分分析法(Two-dimensional Principal Component Analysis,2DPCA)降维预处理,使用光流法对降维后图像提取运动特征;然后使用Radon变换算法对光流图像进行处理,得到对应微表情的特征值和特征图像;最后使用支持向量机进行微表情分类识别。实验结果表明,使用Radon变换后得到的微表情特征图像得到了较好的识别效果,在微表情数据集CASME和CASMEⅡ上识别率分别为81. 48%和82. 17%,通过与选取的其他方法对比说明了该方法具有更好的识别性能。

Micro-expressions are emotional expressions that people try to control and conceal in high-intensity environments different from their usual living environments. They are also instantaneous facial expressions that people do not realize. Micro-expressions have short duration and weak intensity. In order to improve their accuracy,a micro-expression recognition algorithm based on Radon transform is proposed.Firstly,gray level normalization,size normalization and two-dimensional principal component analysis( 2DPCA) dimension reduction preprocessing are carried out for video sequences in the database. After dimension reduction,motion features are extracted by optical flow method. Then,Radon transform algorithm is used to process the flow image,and the corresponding features and feature images of micro-expressions are obtained. Finally,support vector machine is used to classify and recognize micro-expressions. The experimental results show that the micro-expression feature image obtained by Radon transform has a good recognition effect. The recognition rates on CASME and CASMEⅡare 81. 48% and 82. 17% respectively. Compared with other methods,the proposed method has better recognition performance.