改进Softmax分类器的深度卷积神经网络及其在人脸识别中的应用

doi:10.12066/j.issn.1007-2861.1831

摘要/Abstract

摘要：

提出了一种有效的特征学习方法, 构建了9层结构的卷积神经网络,利用Softmax回归算法进行人脸分类识别.卷积神经网络在输出层利用改进的Softmax进行分类,在隐藏层采用修正线性单元作为激活函数,并在网络中加入局部响应归一化处理, 有效避免了梯度消失问题.利用大量人脸图像数据对网络进行预训练, 得到较好的网络初始权重.在针对YALE, FERET, LFW-A等人脸数据库进行人脸识别实验中,与现有的几种人脸识别方法SDAEs, RRC, MPCRC, CRC,SRC等进行对比表明, 该方法在各人脸数据库的识别中均得到较高的识别率.

关键词: 卷积神经网络, 线性修正单元, 局部响应归一化, 人脸识别

Abstract:

An effective learning method by constructing a 9-layer convolution neural network structure, using the softmax regression for human face classification and identification is proposed. Taking advantage of the improved softmax classifier in the output layer, using a rectified linear unit as the activation function in the hidden layer and local response normalization processes in the network, the problem of vanishing gradient is well solved. Pre-trained by using a large number of face images, proper network initial weights are obtained. Experiments based on three face databases YALE, FERET and LFW-A, demonstrate that the proposed method approaches the highest recognition rate compared with other methods such as SDAEs, RRC, MPCRC, CRC and SRC.

Key words: convolution neural network, rectified linear unit, local response normalization, face recognition

中图分类号:

TP183

冉鹏, 王灵, 李昕, 刘鹏伟. 改进Softmax分类器的深度卷积神经网络及其在人脸识别中的应用[J]. 上海大学学报(自然科学版), 2018, 24(3): 352-366.

RAN Peng, WANG Ling, LI Xin, LIU Pengwei. Improved softmax classifier for deep convolution neural networks and its application in face recognition[J]. Journal of Shanghai University（Natural Science Edition）, 2018, 24(3): 352-366.

图/表 14

图1

图2

图3

图4

表1

图5

表2

表3

表4

表5

表6

表7

表8

表9

参考文献 27

[1]	常亮, 邓小明, 周明全, 等. 图像理解中的卷积神经网络[J]. 自动化学报, 2016,42(9):1300-1312.
[2]	王晓刚. 图像识别中的深度学习[J]. 中国计算机学会通讯, 2015,11(8):15-23.
[3]	Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J]. Science, 2006,313:504-507. doi: 10.1126/science.1127647
[4]	孙志远, 鲁成祥, 史忠植, 等. 深度学习研究与进展[J]. 计算机科学, 2016,43(2):1-8.
[5]	Sainath T N, Kingsbury B, Saon G, et al. Deep convolutional neural networks for large-scale speech tasks[J]. Neural Networks, 2015,64:39-48. pmid: 25439765
[6]	王山海, 景新幸, 杨海燕. 基于深度学习神经网络的孤立词语音识别的研究[J]. 计算机应用研究, 2015,32(8):2289-2298.
[7]	Le Q V, Mikolov T. Distributed representations of sentences and documents [C]// International Conference on Machine Learning. 2014: 1188-1196.
[8]	王伟凝, 王励, 赵明权, 等. 基于并行深度卷积神经网络的图像美感分类[J]. 自动化学报, 2016,42(6):904-914.
[9]	Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks [C]// 2012 Advances in Neural Information Processing Systems (NIPS2012). 2012: 1097-1105.
[10]	杜骞. 深度学习在图像语义分类中的应用[D]. 武汉: 华中师范大学, 2014.
[11]	芮挺, 费建超, 周遊, 等. 基于深度卷积神经网络的行人检测[J]. 计算机工程与应用, 2016,52(13):162-166.
[12]	Schroff F, Kalenichenko D, Philbin J. FaceNet: a unified embedding for face recognition and clustering [C]// 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR2015). 2015: 815-823.
[13]	Huang G B, Ramesh M, Berg T, et al. Labeled faces in the wild: a database for studying face recognition in unconstrained environments[M]. Amherst: University of Massachusetts, 2007.
[14]	Sun Y, Wang X, Tang X. Deep learning face representation from predicting 10 000 classes [C]// 2014 IEEE Conference on Computer Vision and Pattern Recognition (CVPR2014). 2014: 1891-1898.
[15]	Taigman Y, Yang M, Ranzato M A, et al. Deepface: closing the gap to human-level performance in face verification [C]// 2014 IEEE Conference on Computer Vision and Pattern Recognition (CVPR2014). 2014: 1701-1708.
[16]	Liu J, Deng Y, Bai T, et al. Targeting ultimate accuracy: face recognition via deep embedding [J]. CoRR, 2015, DOI: http://dblp.uni-trier.de/db/journals/corr/corr1506.html$\#$LiuDBH15.
[17]	柴瑞敏, 曹振基. 基于改进的稀疏深度信念网络的人脸识别方法[J]. 计算机应用研究, 2015,32(7):2179-2183.
[18]	Belhumeur P M, Hespanha J P, Kriegman D J. Eigenfaces vs. fisherfaces: recognition using class specific linear projection [C]// IEEE Transactions on Pattern Analysis and Machine Intelligence. 2002: 711-720.
[19]	Phillips P J, Moon H, Rizvi S A, et al. The FERET evaluation methodology for face recognition algorithms [C]// IEEE Transactions on Pattern Analysis and Machine Intelligence. 2000: 1090-1104.
[20]	LeCun Y, Bottou L, Bengio Y, et al. Gradient-based learning applied to document recognition[J]. Proceeding of the IEEE, 1998,86(11):2278-2324.
[21]	Hochreiter S, Bengio Y, Frasconi P, et al. Gradient flow in recurrent nets: the difficulty of learning long-term dependencies [M]. Piscataway: IEEE Press, 2001: 1-15.
[22]	Ren S Q, Cao X D, Wei Y C, et al. Face alignment at 3000 FPS via regressing local binary features [C]// 2014 IEEE Conference on Computer Vision and Pattern Recognition (CVPR2014). 2014: 1685-1692.
[23]	Zhu P, Zhang L, Hu Q, et al. Multi-scale patch based collaborative representation for face recognition with margin distribution optimization [C]// 2012 European Conference on Computer Vision (ECCV2012). 2012: 822-835.
[24]	Pang S C, Yu Z Z. Face recognition: a novel deep learning approach[J]. Journal of Optical Technology, 2015,82(4):237-245.
[25]	Yang M, Zhang L, Yang J, et al. Regularized robust coding for face recognition[J]. IEEE Transactions on Image Processing, 2013,22(5):1753-1766. doi: 10.1109/TIP.2012.2235849 pmid: 23269753
[26]	Zhang L, Yang M, Feng X. Sparse representation or collaborative representation: which helps face recognition [C]// 2011 IEEE International Conference on Computer Vision (ICCV2011). 2011: 471-478.
[27]	Wright J, Yang A, Ganesh A, et al. Robust face recognition via sparse representation [C]// IEEE Transactions on Pattern Analysis and Machine Intelligence. 2009: 210-212.