摘要
提出了一种基于二维离散小波-离散余弦变换-支持向量机(DW T-DCT-SVM)的面部表情识别算法。该算法先利用DW T在不明显损失图像信息的基础上对表情图像进行变换,变换后的图像数据量大大减少。再利用DCT提取代表原图像绝大部分能量的数据作为表情特征矢量,最后利用SVM来识别。实验表明:本算法提取的500个数据长度的表情矢量在一定条件下能较准确地识别出通用的6种表情,但是泛化性能较差。
An effective method for the facial expression recognition is presented based on the discrete wavelet transform （DWT） and the discrete cosine transform（DCT）. Firstly, the wave packet decomposition and the discrete cosine transform are carried out on the generalized facial images. Then, expression feature vectors are given, because DCT can remove the correlation and accumulate the energy from the low frequency data generated by the discrete wavelet packet decomposition. Finally, SVM is used to make a classification. Experimental results indicate that the method can work very well.
出处
《数据采集与处理》
CSCD
北大核心
2006年第1期64-68,共5页
Journal of Data Acquisition and Processing
基金
教育部"面向21世纪教育振兴行动计划"资助项目
关键词
表情识别
离散小波变换
离散余弦变换
支撑向量机
facial expression recognition
discrete wavelet transform
discrete cosine transform
support vector machine
