摘要
针对包含表情信息的静态图像,提出基于Gabor小波和SVM的人脸表情识别算法。根据先验知识,并使用形态学和积分投影相结合定位眉毛眼睛区域,采用模板内计算均值定位嘴巴区域,自动分割出表情子区域。对分割出的表情子区域进行Gabor小波特征提取,在利用Fisher线性判别对特征进行降维,去除冗余和相关。利用支持向量机对人脸表情进行分类。用该算法在日本表情数据库上进行测试,获得了较高的识别准确率。证明了该算法的有效性。
A facial recognition algorithm based on Gabor wavelet and SVM is proposed in allusion to static image containing expression Information.The mathematical morphology combined with projection is adopted to locate the brow and eye region,and the calculating mean value in template is employed to locate the mouth region,which can segment the expression sub-regions automatically.The features of the expression sub-regions are extracted by Gabor wavelet transformation and then effective Gabor expression features are selected by Fisher linear discriminate（FLD） to deduce the dimension and redundancy of the features.The features are sent to support vector machine（SVM） to classify the different expressions.The algorithm was tested in Japanese female expression database.It can get a high precision of recognition.The feasibility of this method was verified by experiments.
出处
《现代电子技术》
2011年第20期1-4,共4页
Modern Electronics Technique
基金
西北工业大学2011年度研究生创业种子基金资助项目（Z2011090）
关键词
GABOR小波变换
表情特征提取
FISHER线性判别
支持向量机
Gabor wavelet transform
expression feature extraction
Fisher linear discriminant（FLD） analysis
support vector machine（SVM）
