首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于小波变换和非负矩阵分解的人脸识别改进方法
引用本文:杨绍华,牛万红,潘晨.基于小波变换和非负矩阵分解的人脸识别改进方法[J].宁夏大学学报(自然科学版),2010,31(2):130-134.
作者姓名:杨绍华  牛万红  潘晨
作者单位:1. 宁夏大学,数学计算机学院,宁夏,银川,750021
2. 宁夏大学远程教育学院,宁夏,银川,750021
3. 中国计量学院信息工程学院,浙江杭州,310018
基金项目:国家自然科学基金资助项目,宁夏自然科学基金资助项目,宁夏大学科研基金资助项目((E)NDZR09-29;,宁夏大学"211工程"学科建设项目 
摘    要:针对NMF在识别人脸图像特征时分解速度慢,基空间不适应欧氏距离度量的缺点,提出了一种基于小波变换和非负矩阵基矩阵正交化的人脸识别方法.利用小波变换对人脸图像进行变换,选择LL分量既能抽取到人脸的实质特征又能有效减小数据维数,降低NMF分解的复杂性.同时对NMF的基矩阵实施正交化变换,在得到的正交基上进行投影.实验结果表明,该方法对光照变化、表情变化和部分遮罩不敏感,识别性能明显提高.

关 键 词:人脸识别  小波变换  非负矩阵分解

An Improved Face Recognition Method Based on Wavelet Transform and Nonnegative Matrix Factorization
Yang Shaohua,Niu Wanhong,Pan Chen.An Improved Face Recognition Method Based on Wavelet Transform and Nonnegative Matrix Factorization[J].Journal of Ningxia University(Natural Science Edition),2010,31(2):130-134.
Authors:Yang Shaohua  Niu Wanhong  Pan Chen
Institution:1.School of Mathematics and Computer Science,Ningxia University,Yinchuan 750021,China; 2.School of Remote Education,Ningxia University,Yinchuan 750021,China; 3.School of Information Engineering,China Jiliang University,Hangzhou 310018,China)
Abstract:Aiming at the lower resolving speed of NMF for face recognition and the disadvantages of space of bases not suitable for Euclid distance measuring,the method of face recognition with the use of wavelet transformation and the orthonormalized bases of nonnegative matrix is put forward.Making use of wavelet transformation to decompose the face images and choosing the LL coefficients,not only central facial features are captured but also computational complexity of NMF is reduced.Also,orthogonal transformation is put into base-matrix of NMF,and training samples and testing samples are projected onto orthogonal base-matrix.The experimental results show that the above-mentioned algorithm performs faster training speed and higher accuracy of recognition against the lower sensitivity to illumination,facial expression and part covering.
Keywords:face recognition  wavelet transform  nonnegative matrix factorization
本文献已被 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号