摘要
针对Gabor特征维数和冗余度较高的缺点,对Gabor小波变换特征进行分块,提取了所有子块局部统计特征,然后使用PCA+LDA方法对这些特征进行选择,最后采用决策树分类法进行了人脸表情识别.实验结果表明:此方法在维数降低的同时,其识别性能比传统的方法更具优势.
Aiming at the deficiency of the high-dimensional and high-redundant Gabor feature vectors,firstly the Gabor wavelet coefficients are divided into blocks and the local statistical features of all Gabor wavelet representation are calculated as the feature vectors,then the sorted PCA plus LDA is used to select and compress the Gabor features,and finally a decision tree classifier is adopted to recognize facial expression. Experimental results show that the method is more effective for both dimension reduction and recognition performance than some traditional methods.
出处
《中南民族大学学报（自然科学版）》
CAS
2010年第1期78-82,共5页
Journal of South-Central University for Nationalities：Natural Science Edition
基金
国家民委自然科学基金资助项目(MZZ04004)
中南民族大学自然科学基金资助项目(YZZ05003)
作者简介
高智勇（1972-），男，博士，副教授，研究方向：图像处理和识别，E-mail：zhiyonggao@mail．scuec．edu．cn
