计算机科学 ›› 2015, Vol. 42 ›› Issue (9): 61-65.doi: 10.11896/j.issn.1002-137X.2015.09.013
王莹,樊鑫,李豪杰,林妙真
WANG Ying, FAN Xin, LI Hao-jie and LIN Miao-zhen
摘要: 在实际的自动人脸识别系统中,输入的识别图像往往在表情、分辨率大小以及姿态方面呈现出多种变化。现在很多方法尝试通过线性或局部线性的映射来寻找由这些变化共享的统一的特征空间。利用由受限玻尔兹曼机(RBM)堆叠成的深度神经网络来发掘这些变化内在的非线性表达。深度网络能够学习高维数据到低维数据的映射关系,并且有助于提高图像分类和识别的性能。同时,为了实现在一个统一的深度框架下同时进行特征提取和识别,在网络的顶层增加了一个监督的回归层。在预训练阶段,通过训练集中不同姿态、不同表情以及不同分辨率的图像对网络进行初始化。在微调阶段,通过网络的输出与标签之间的差 并利用标准反向传播的方法 对模型的参数空间进行调整。在测试阶段,从测试库中随机选择一幅图像,获得统一空间下的特征向量。通过与参考图像库中的所有特征向量进行对比,利用最近邻域的方法识别人脸身份。在具有丰富表情以及大姿态变化的CMU-PIE人脸数据库上进行了全面的实验,结果表明,提出的方法取得了比最新的局域线性映射(或局部线性)的人脸识别方法更高的识别率。
[1] Huang H,Zeng X.Super-resolution method for multi-view face recognition from a single image per person using nonlinear mappings on coherent features [J].IEEE Signal Processing Letters,2012,19(4):195-198 [2] Zou W,Yuen P.Very low resolution face recognition problem[C]∥IEEE Transactions on Image Processing,2012 [3] Zhang X,Gao Y.Face recognition across pose.A review [J].Pattern Recognition,2009,42(11):2876-2896 [4] Huang H,He H.Super-resolution method for face recognition using nonlinear mappings on coherent features [J].IEEE Transactions on Neural Networks,2011,22(1):121-130 [5] Li B,Chang H,Shan S,et al.Low-resolution face recognition via coupled locality preserving mappings [J].IEEE Signal Proces-sing Letters,2010,17(1):20-23 [6] Hinton G E,Salakhutdinov R R.Reducing the dimensionality of data with neural networks [J].Science,2006,313(5786):504-507 [7] Salakhutdinov R,Hinton G E.Learning a nonlinear embedding by preserving class neighbourhood structure [J].International Conference on Artificial Intelligence and Statistics,2007,3:412-419 [8] Zhou S,Chen Q,Wang X.Discriminative deep belief networks for image classification [C]∥ICIP.2010:1561-1564 [9] Mohamed A R,Dahl G,Hinton G.Deep belief networks forphone recognition [C]∥NIPS Workshop on Deep Learning for Speech Recognition and Related Applications.2009 [10] Sim T,Baker S,Bsat M.The CMU pose,illumination,and expression(pie) database [C]∥Proceedings of International Conference on Automatic Face and Gesture Recognition.2002:46-51 [11] Hinton G E,Osindero S,Teh Y-W.A fast learning algorithm for deep belief nets [J].Neural Computation,2006,18(7):1527-1554 [12] Huang G B,Lee H,Learned-Miller E.Hierarchical representations for face verification with convolutional deep belief networks [C]∥Proceedings of International Conference on Computer Vision and Pattern Recognition(CVPR).2012:2518-2525 [13] 何加浪,张宏.神经网络在软件多故障定位中的应用研究[J].计算机研究与发展,2013,50(3):619-625 He Jia-lang,Zhang Hong.Application of artificial neural network in software multi-faults location [J].Journal of Computer Research and Development,2013,50(3):619-625 [14] 周旭东,陈晓红,陈松灿.半配对半监督场景下的低分辨率人脸识别[J].计算机研究与发展,2012,49(11):2328-2333 Zhou Xu-dong,Chen Xiao-hong,Chen Song-can.Low-Resolution face recognition in semi-paired and semi-supervised scenario [J] Journal of Computer Research and Development,2012,49(11):2328-2333 [15] Andrew G,Arora R,Bilmes J,et al.Deep Canonical Correlation Analysis[J].JMLR W&CP,2013,28(3):1247-1255 [16] Le Q V,Ngiam Ji-quan,Chen Zheng-hao,et al.Tiled convolutional neural networks [C]∥NIPS.2010 [17] Zhu Zhen-yao,Luo Ping,Wang Xiao-gang,et al.Deep Learning Identity-Preserving Face Space [C]∥ICCV 2013.2013:113-120 [18] Kan Mei-na,Shan Shi-guang,Chang Hong,et al.Stacked Progressive Auto-Encoders for Face Recognition Across Poses[C]∥CVPR 2014.2014:1883-1890 [19] 胡石,李光辉,卢文伟,等.基于神经网络的无线传感器网络异常数据检测方法 [J].计算机科学,2014,41(11A):208-211 Hu Shi,Li Guang-hui,Lu Wen-wei et al.Outlier Detection Methods based on Neural Network in Wireless Sensor Networks [J].Computer Science,2014,41(11A):208-211 [20] 刘逻,郭力,红肖辉,等.基于参数动态调整的动态模糊神经网络的软件可靠性增长模型 [J].计算机科学,2013,40(2):186-190 Liu Lu,Guo Li,Hong Xiao-hui,et al.Software reliability growth model based on dynamic fuzzy neural network with parameters dynamic adjustment[J].Computer Science,2013,40(2):186-190 |
No related articles found! |
|