基于奇异值分解和小波神经网络的人脸识别
|
摘要
针对人脸识别在有遮挡、表情变化和光照变化引起的鲁棒性变差问题,以及传统人工神经网络用于人脸识别时存在的维数灾难问题,提出一种分块奇异值分解和小波神经网络结合的人脸识别算法。首先,将人脸图像进行分块,获得图片局部的奇异值,并将其按一定顺序排列得到人脸的特征向量;然后,运用加入动量项的改进小波神经网络进行人脸图像分类识别;最后,在Matlab环境下利用ORL和YALE人脸图像数据库进行仿真实验,并且在GUI图形用户界面上进行验证。实验结果表明,该算法实现简单,识别率高,对光照、遮挡、表情等变化有很好的鲁棒性,具有很大的使用价值。
In allusion to the poor robustness problem caused by shielding and variations of face expression and illumination during face recognition,and the dimension disaster problem existing in the traditional artificial neural network used for human face recognition,a human face recognition algorithm combining blocked singular value decomposition with wavelet neural network is proposed. The human face images are blocked to obtain local singular values of images,which are ranked in a certain order to obtain feature vectors of human faces. The human face image classification and recognition are conducted by using the improved wavelet neural network with momentum flux added. The simulation experiment was carried out with the ORL and YALE human face image databases in the Matlab environment. The algorithm was verified on the GUI graphical user interface.The experimental results show that the algorithm is simple to be implemented,and has a high recognition rate and good robustness to variations of illumination,shielding and face expression,which is of great use value.
In allusion to the poor robustness problem caused by shielding and variations of face expression and illumination during face recognition,and the dimension disaster problem existing in the traditional artificial neural network used for human face recognition,a human face recognition algorithm combining blocked singular value decomposition with wavelet neural network is proposed. The human face images are blocked to obtain local singular values of images,which are ranked in a certain order to obtain feature vectors of human faces. The human face image classification and recognition are conducted by using the improved wavelet neural network with momentum flux added. The simulation experiment was carried out with the ORL and YALE human face image databases in the Matlab environment. The algorithm was verified on the GUI graphical user interface.The experimental results show that the algorithm is simple to be implemented,and has a high recognition rate and good robustness to variations of illumination,shielding and face expression,which is of great use value.
引文
[1] BECKER B C,ORTIZ E G. Evaluating open-universe face identification on the Web[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition Workshops. Portland:IEEE,2013:904-911.
[2]孔令美,汤庸.基于小波变换和小波神经网络的3D遮挡人脸识别方法[J].湘潭大学自然科学学报,2015,37(4):82-86.KONG Lingmei,TANG Yong. A method for 3D occlusion face recognition based on wavelet transform and wavelet neural network[J]. Natural science journal of Xiangtan University,2015,37(4):82-86.
[3] CAO D,YANG B. An improved face recognition algorithm based on SVD[C]//Proceedings of 2nd International Conference on Computer and Automation Engineering. Singapore:IEEE,2010:109-112.
[4]马园园,张登银.基于SVD的两步人脸识别方法[J].计算机技术与发展,2018,28(1):95-99.MA Yuanyuan,ZHANG Dengyin. A two-step face recognition method based on SVD[J]. Computer technology and development,2018,28(1):95-99.
[5]董超,徐宁,Kim Batselier,等.基于TTr1SVD的张量奇异值分解及其在人脸识别上的应用[J].计算机应用研究,2018,35(1):287-291.DONG Chao,XU Ning,BATSELIER K,et al. Tensor singular value decompositions based on TTr1SVD and their applications to face recognition problem[J]. Application research of computers,2018,35(1):287-291.
[6] ZHANG Q, BENVENISTE A. Wavelet networks[J]. IEEE transactions on neural networks,1992,3(6):889-898.
[7] YU M,WANG H,LAN Ting. The BP neural network for improvement of classification accuracy in remote sensing image[J]. International journal of environmental protection and policy,2016,4(3):93-97.
[8] ZHANG Guiying,ZOU Wenbin,ZHANG Xianjie,et al. Singular value decomposition based sample diversity and adaptive weighted fusion for face recognition[J]. Digital signal processing,2017,62:150-156.
[9]张慈祥,刘辉,强振平.基于稀疏表示和奇异值分解的人脸识别[J].计算机应用,2013,33(z1):233-235.ZHANG Cixiang,LIU Hui,QIANG Zhenping. Face recognition based on image sparse representation and singular value decomposition[J]. Journal of computer applications,2013,33(S1):233-235.
[10]姜友谊.小波神经网络在人脸识别中的应用[J].西安科技大学学报,2012,32(5):652-657.JIANG Youyi. Face recognition based on wavelet neural network[J]. Journal of Xi’an University of Science and Technology,2012,32(5):652-657.
[11] GUI G,MEHBODNIYA A,ADACHI F. Sparse LMS/F algorithms with application to adaptive system identification[J].Wireless communications&mobile computing, 2015, 15(12):1649-1658.
[12]何正风,孙亚民.基于奇异值分解和径向基函数神经网络的人脸识别算法研究[J].计算机科学,2012,39(z1):566-569.HE Zhengfeng,SUN Yamin. Research on face recognition algorithm based on SVD and RBF neural network[J]. Computer science,2012,39(S1):566-569.
[2]孔令美,汤庸.基于小波变换和小波神经网络的3D遮挡人脸识别方法[J].湘潭大学自然科学学报,2015,37(4):82-86.KONG Lingmei,TANG Yong. A method for 3D occlusion face recognition based on wavelet transform and wavelet neural network[J]. Natural science journal of Xiangtan University,2015,37(4):82-86.
[3] CAO D,YANG B. An improved face recognition algorithm based on SVD[C]//Proceedings of 2nd International Conference on Computer and Automation Engineering. Singapore:IEEE,2010:109-112.
[4]马园园,张登银.基于SVD的两步人脸识别方法[J].计算机技术与发展,2018,28(1):95-99.MA Yuanyuan,ZHANG Dengyin. A two-step face recognition method based on SVD[J]. Computer technology and development,2018,28(1):95-99.
[5]董超,徐宁,Kim Batselier,等.基于TTr1SVD的张量奇异值分解及其在人脸识别上的应用[J].计算机应用研究,2018,35(1):287-291.DONG Chao,XU Ning,BATSELIER K,et al. Tensor singular value decompositions based on TTr1SVD and their applications to face recognition problem[J]. Application research of computers,2018,35(1):287-291.
[6] ZHANG Q, BENVENISTE A. Wavelet networks[J]. IEEE transactions on neural networks,1992,3(6):889-898.
[7] YU M,WANG H,LAN Ting. The BP neural network for improvement of classification accuracy in remote sensing image[J]. International journal of environmental protection and policy,2016,4(3):93-97.
[8] ZHANG Guiying,ZOU Wenbin,ZHANG Xianjie,et al. Singular value decomposition based sample diversity and adaptive weighted fusion for face recognition[J]. Digital signal processing,2017,62:150-156.
[9]张慈祥,刘辉,强振平.基于稀疏表示和奇异值分解的人脸识别[J].计算机应用,2013,33(z1):233-235.ZHANG Cixiang,LIU Hui,QIANG Zhenping. Face recognition based on image sparse representation and singular value decomposition[J]. Journal of computer applications,2013,33(S1):233-235.
[10]姜友谊.小波神经网络在人脸识别中的应用[J].西安科技大学学报,2012,32(5):652-657.JIANG Youyi. Face recognition based on wavelet neural network[J]. Journal of Xi’an University of Science and Technology,2012,32(5):652-657.
[11] GUI G,MEHBODNIYA A,ADACHI F. Sparse LMS/F algorithms with application to adaptive system identification[J].Wireless communications&mobile computing, 2015, 15(12):1649-1658.
[12]何正风,孙亚民.基于奇异值分解和径向基函数神经网络的人脸识别算法研究[J].计算机科学,2012,39(z1):566-569.HE Zhengfeng,SUN Yamin. Research on face recognition algorithm based on SVD and RBF neural network[J]. Computer science,2012,39(S1):566-569.