查询结果:   郑翔,鲜敏,马勇.增强Kernel学习优化最大边缘投影的人脸识别[J].计算机应用与软件,2015,32(9):314 - 318.
中文标题
增强Kernel学习优化最大边缘投影的人脸识别
发表栏目
信息技术交流
摘要点击数
832
英文标题
MAXIMUM MARGIN PROJECTION FOR FACE RECOGNITION OPTIMISED BY ENHANCED KERNEL LEARNING
作 者
郑翔 鲜敏 马勇 Zheng Xiang Xian Min Ma Yong
作者单位
四川工程职业技术学院计算机科学技术系 四川 德阳 618000     
英文单位
Department of Computer Science and Technology, Sichuan Engineering Technical College, Deyang 618000, Sichuan, China     
关键词
人脸识别 最大边缘投影 支持向量机 增强核学习 特征向量选择
Keywords
Face recognition Maximum margin projection Support vector machine Enhanced kernel learning Feature vector selection
基金项目
作者资料
郑翔,讲师,主研领域:模式识别,图像处理。鲜敏,讲师。马勇,副教授。 。
文章摘要
针对传统的流形学习算法通常只考虑样本类内几何结构而忽略类间判别信息的问题,提出一种基于增强核学习的最大边缘投影(MMP)算法。首先使用基于增强核学习非线性扩展的MMP采集人脸图像的非线性结构;然后利用核变换技术加强原始输入核函数的判别能力,并且借助于特征向量选择算法改善算法的计算效率;最后,利用基于乘性规则训练的支持向量机完成人脸的识别。在Yale、ORL、PIE三大通用人脸数据库的组合数据集及AR上的实验验证了该算法的有效性。实验结果表明,相比其他几种核学习算法,该算法取得了更好的识别效果。
Abstract
For the problem that traditional manifold learning methods usually consider intra-class geometry structure only but ignore the discriminative information of inter-classes, we propose an enhanced kernel learning-based maximum margin projection (MMP) algorithm. Firstly, we use MMP nonlinearly extended by enhanced kernel learning to collect the nonlinear structure of face image. Then, we use kernel transformation technology to enhance the discriminant ability of original inputted kernel function, and improve the computation efficiency of the proposed algorithm by feature vector selection algorithm. Finally, we use support vector machine trained by multiplicative rules to finish the face recognition. The effectiveness of the proposed algorithm is verified by the experiments on AR and the combination datasets of three common face databases Yale, ORL and PIE. Experimental results show that the proposed algorithm has better recognition efficiency comparing with several other advanced approaches based on kernel learning.
下载PDF全文