摘要
对于人脸识别问题 ,基于 K- L 变换对人脸图像进行特征参数的提取 ;并采用支持向量机进行分类 .由于支持向量机本身是一个两类问题的判别方法 ,在处理多类问题时 ,提出了一种基于支持向量机组的淘汰法 ,这种方法考虑到了各判别函数的 VC置信范围的差异 ,同时利用判别函数间的冗余来降低识别误差 .针对 ORL 人脸库和自建的人脸库的识别结果表明 ,基于 SVM的识别方法在特征参数个数的选取、识别效果、识别时间等方面都有其独到的优越性 .
In face recognition, Karhunen-Loeve transform is employed to get the representation basis of face image set, and support vector machines are used to classify. Support Vector Machines (SVM) are classifiers which have demonstrated high generalization capabilities. SVM group incorporated with the elimination strategy is proposed to deal with the multi-class face recognition problem. For considering the differences among the confidence interval about the SVM and the redundancy of each SVM, a higher recognition rate can be reached. The face recognition experiment with the two face databases shows that the proposed method has reached a higher recognition rate and a reasonable time cost.
出处
《小型微型计算机系统》
CSCD
北大核心
2004年第1期139-142,共4页
Journal of Chinese Computer Systems
关键词
人脸识别
支持向量机
淘汰法
本征脸
face recognition
support vector machines
elimination strategy
eigenfaces
