基于Gabor小波变换的人脸表情识别技术研究  

Facial expression recognition based on Gabor wavelet transform

作　　者：张余敬[1] 常丹华[1] 刘宇[1] 张昆[1] 孙志华[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《计算机测量与控制》2010年第4期906-908,911,共4页Computer Measurement &Control

摘　　要：为了使人机交互得到更好的研究,提出了一种基于Gabor小波变换的人脸表情识别的新方法;首先对图像进行预处理以提高后续处理的准确度,预处理包括确定纯的人脸表情区域,尺寸和灰度归一化,然后对表情子区域进行Gabor小波变换,提取表情特征矢量,进而构建表情弹性图,最后用模板匹配的方法来识别图像的人脸表情;经过实验,发现Gabor小波变换提取特征时受光照影响比较小,该系统不仅具有很好的鲁棒性,并且速度快,识别率高。To give a better research of human-computer interaction, the new method of facial expression recognition based on Gabor wavelets transform was discussed. Image preprocessors were executed firstly to improve the accuracy of follow-up processing. Preprocessors including the identification of pure Face Facial Expression region, size and gray-scale normalized. Secondly, expression feature vectors of the expression sub-regions were extracted by Gabor wavelet transformation to form expression elastic graph. Finally, template matching methods was used to identify the facial expression images. Experiment shows that expression features can be extracted efectively based on Gabor wavelet transformation, which is insensitive to illumination variety, the system has good robust and speed, and high recognition accuracy.

关 键 词：表情识别 预处理 GABOR小波变换 特征提取 弹性模板匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]