摘要
关于人脸识别问题,采用一种基于独立分量分析进行特征提取和支持向量机实现多分类的人脸识别新方法。根据支持向量机理论,为提高对人脸的识别率,提出环形对称划分的支持向量机多分类算法。算法将多类问题的类别环形排列,依次进行对称划分构造纠错编码输出矩阵;根据求得的纠错编码输出矩阵,用解码函数求解待求样本的类别。对于人脸识别问题,利用独立分量分析方法构造人脸的特征脸空间,在特征脸空间运用算法进行人脸识别,在人脸数据库上的仿真结果表明,算法能有效地完成人脸识别任务。
A new method of face recognition based on Support Vector Machine（SVM） is proposed.According to the theory of SVM,a novel method is proposed for solving multi-class classification problem based on loop-symmetrical division SVM.This algorithm loop-arranges and symmetrical divides classes in multi-class problem,and constructs error-correcting codes matrix.Based on the error-correcting codes matrix,the class of unknown sample is found by decoded function.For face recognition,eigenface space is constructed depending on Independent Component Analysis（ICA）.And in eigenface space,this method is applied to recognize and classify face images.Experimental results with the face database demonstrate the efficiency of the new algorithm.
出处
《计算机仿真》
CSCD
北大核心
2011年第2期280-283,共4页
Computer Simulation
基金
国家自然科学基金(10875027)
