摘要
本文提出一种人脸特征复合选择的新方法。首先对原始图像进行小波2阶分解和KPCA进行特征提取,然后将获得的特征进行SVM训练,经过GSFS反复选择具有最小间隔的支持向量作为最佳特征组合,最后输入线性SVM分类器进行分类。实验报告了本方法在UMIST及IITL人脸数据库上的应用,并对特征选择前后的分类能力及速度进行了比较,结果显示经过本方法的特征选择后,人脸识别能力有所提高,分类速度明显加快。
In this paper, a novel feature selection method is proposed for face recognition. The proposed method can be described as follows, first wavelet transformation is used for decomposing initial face features to obtain 2nd order low-frequency information and by using KPCA the non-linear features are extracted from the obtained low-frequency information. Secondly, the extracted non-linear features are trained by SVM for selecting the optimal features with minimum margin support vectors and the features are divided into some single feature sets. Lastly, the single feature sets are trained separately by SVM to obtain the best feature set through GSFS. In this way, the dimensionality of the initial features can be reduced dramatically. Experimental results on UMIST and IITL face databases indicated the effectiveness of the proposed method.
出处
《电子测量与仪器学报》
CSCD
2006年第2期16-20,共5页
Journal of Electronic Measurement and Instrumentation
基金
国家教育部科学研究重点项目(编号:02057)
重庆市自然科学基金重点研究项目(编号:CSTC2005BA2002)
重庆市自然科学基金项目(编号:CSTC2005BB2181)。
关键词
人脸识别
小波变换
核主元分析法
支持向量机
广义顺序前进法
face recognition, wavelet transform, kernel principal component analysis, support vctor machine,generalized sequential forward selection.