正在加载图片...
第3期 马晓,等:基于深度学习特征的稀疏表示的人脸识别方法 ·285. topy continuation for sparse signal representation C//Pro- 5 结束语 ceedings of the IEEE International Conference on Acoustics, 本文针对较大类内变化干扰下的人脸识别问 Speech,and Signal Processing,Philadelphia,USA,2005: 题,提出了基于深度学习特征稀疏表示的人脸识别 733-736. 方法SRDLFC。本文充分分析论证了基于深度学习 [10]KOH K,KIM S J,BOYD S P.An interior-point method for large-scale 11-regularized logistic regression J.Jour- 所提取的人脸特征基本满足线性子空间假设,并具 nal of machine learning research,2007,8(8):1519- 有较好的可分性、可迁移性及对类内变化的不变性。 1555. 本文将基于深度学习的特征应用到稀疏表示的分类 [11 LIU Y,WU F,ZHANG Z.Sparse representation using 框架中,充分发挥两者优点,实现识别率的提升。本 nonnegative curds and whey [C]//Proceedings of the 文提出的SRDLFC算法,可以有效地应对光照、姿 IEEE International Conference on Computer Vision and 态、表情、遮挡等类内变化带来的干扰,且在小样本 Pattern Recognition.San Francisco,USA,2010,119 问题中具有较大的优势。未来的研究工作将进一步 (5):3578-3585. 对深度学习特征进行研究分析,通过改进网络结构 [12]GAO S,TSANG I W,Chia L,et al.Local features are not 和损失函数,使网络所得特征更加满足线性子空间 lonely-laplacian sparse coding for image classification 约束,进一步提升SRDLFC的识别效果,并尝试做一 [C]//Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition,San Francis- 些理论上的推导工作。满足线性子空间约束,进一 co,USA,2010,23(3):3555-3561. 步提升SRDLFC的识别效果,并尝试做一些理论上 [13]FISER J,BERKES P,WHITE B.No evidence for active 的推导工作。 sparsification in the visual cortex[C]//Advances in Neu- 参考文献: ral Information Processing Systems,Vancouver,B.C., Canada,2009:108-116. [1]DONOHO D L.Compressed sensing[J].Information theory. [14]ZHANG D,YANG M,FENG X.Sparse representation or 2006,52(4):1289-1306. collaborative representation:which helps face recognition? [2]LEE K C,HO J,KRIEGMAN D.Acquiring linear sub- [C]//Proceedings of the IEEE International Conference spaces for face recognition under variable lighting[J].Pat- on Computer Vision.Barcelona,Spain,2011,6669(5): tern analysis and machine intelligence,2005,27(5):684- 471-478. 698. [15]DENG W,HU J,GUO J.Extended SRC:undersampled [3]NASEEM I,TOGNERI R,BENNAMOUN M.Linear regres- face recognition via intraclass variant dictionary[J].Pat- sion for face recognition[].Pattern analysis and machine tern analysis and machine intelligence,IEEE transactions intelligence,2010,32(11):2106-2112. om,2012,34(9):1864-1870. [4]WRIGHT J.YANG A Y,GANESH A,et al.Robust face [16]GUO J.In defense of sparsity based face recognition[C]/ recognition via sparse representation[J].Pattern analysis Proceedings of the IEEE International Conference on Com- and machine intelligence,2009,31(2):210-227. puter Vision.Sydney,NSW,Australia,2013,9(4):399- [5]AHARON M,ELAD M,BRUCKSTEIN A.K-SVD:an al- 406. gorithm for designing overcomplete dictionaries for sparse [17]SU Y,SHAN S,CHEN X,et al.Adaptive generic learn- representation [J].Signal processing,2006,54(11): ing for face recognition from a single sample per person 4311-4322. [C]//Proceedings of the IEEE International Conference [6]YANG M,ZHANG L,YANG J,et al.Metaface learning for on Computer Vision and Pattern Recognition.San Francis- sparse representation based face recognition[C]//Proceed- co,USA,2010:2699-2706. ings of the IEEE International Conference on Image Process- [18]WEI C,WANG Y-F.Learning auxiliary dictionaries for ing,Hong Kong,China,.2010,119(5):1601-1604. undersampled face recognition [C]//Proceedings of the [7]YANG M,ZHANG D,FENG X,et al.Fisher discrimina- IEEE International Conference on Multimedia and Expo. tion dictionary learning for sparse representation[C]//Pro- San Jose,California,USA,2013,2013:1-6. ceedings of the IEEE International Conference on Computer [19]WANG X,TANG X.Unified subspace analysis for face Vision.Barcelona,Spain,2011,24(4):543-550. recognition[C]//null.Proceedings of the IEEE Interna- [8]YANG J,ZHANG Y.Alternating direction algorithms for 11- tional Conference on Computer Vision.Nice,France, problems in compressive sensing J].arXiv:0912.1185, 2003:679-686. 2009,(1):250-278. [20]LECUN Y L,BOTTOU L,BENGIO Y,et al.Gradient- [9]UJDAT D M,MALIOUTOV D M,CETIN M,et al.Homo- based learning applied to document recognition.Proc IEEE5 结束语 本文针对较大类内变化干扰下的人脸识别问 题,提出了基于深度学习特征稀疏表示的人脸识别 方法 SRDLFC。 本文充分分析论证了基于深度学习 所提取的人脸特征基本满足线性子空间假设,并具 有较好的可分性、可迁移性及对类内变化的不变性。 本文将基于深度学习的特征应用到稀疏表示的分类 框架中,充分发挥两者优点,实现识别率的提升。 本 文提出的 SRDLFC 算法,可以有效地应对光照、姿 态、表情、遮挡等类内变化带来的干扰,且在小样本 问题中具有较大的优势。 未来的研究工作将进一步 对深度学习特征进行研究分析,通过改进网络结构 和损失函数,使网络所得特征更加满足线性子空间 约束,进一步提升 SRDLFC 的识别效果,并尝试做一 些理论上的推导工作。 满足线性子空间约束,进一 步提升 SRDLFC 的识别效果,并尝试做一些理论上 的推导工作。 参考文献: [1]DONOHO D L. Compressed sensing[J]. Information theory, 2006, 52(4): 1289⁃1306. [2] LEE K C, HO J, KRIEGMAN D. Acquiring linear sub⁃ spaces for face recognition under variable lighting[ J]. Pat⁃ tern analysis and machine intelligence, 2005, 27(5): 684⁃ 698. [3]NASEEM I, TOGNERI R, BENNAMOUN M. Linear regres⁃ sion for face recognition[ J]. Pattern analysis and machine intelligence, 2010, 32(11): 2106⁃2112. [4]WRIGHT J, YANG A Y, GANESH A, et al. Robust face recognition via sparse representation [ J]. Pattern analysis and machine intelligence, 2009, 31(2): 210⁃227. [5]AHARON M, ELAD M, BRUCKSTEIN A. K⁃SVD: an al⁃ gorithm for designing overcomplete dictionaries for sparse representation [ J ]. Signal processing, 2006, 54 ( 11 ): 4311⁃4322. [6]YANG M, ZHANG L, YANG J, et al. Metaface learning for sparse representation based face recognition[C] / / Proceed⁃ ings of the IEEE International Conference on Image Process⁃ ing, Hong Kong, China, 2010, 119(5):1601⁃1604. [7]YANG M, ZHANG D, FENG X, et al. Fisher discrimina⁃ tion dictionary learning for sparse representation[C] / / Pro⁃ ceedings of the IEEE International Conference on Computer Vision. Barcelona, Spain, 2011, 24(4):543⁃550. [8]YANG J, ZHANG Y. Alternating direction algorithms for l1⁃ problems in compressive sensing [ J]. arXiv: 0912. 1185, 2009, (1):250⁃278. [9]UJDAT D M, MALIOUTOV D M, ÇETIN M, et al. Homo⁃ topy continuation for sparse signal representation[C] / / Pro⁃ ceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing,Philadelphia, USA, 2005: 733⁃736. [10]KOH K, KIM S J, BOYD S P. An interior⁃point method for large⁃scale l1⁃regularized logistic regression[ J]. Jour⁃ nal of machine learning research, 2007, 8 ( 8): 1519⁃ 1555. [11] LIU Y, WU F, ZHANG Z. Sparse representation using nonnegative curds and whey [ C] / / Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition. San Francisco, USA, 2010, 119 (5):3578⁃3585. [12]GAO S, TSANG I W, Chia L, et al. Local features are not lonely⁃laplacian sparse coding for image classification [C] / / Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition, San Francis⁃ co, USA, 2010, 23(3):3555⁃3561. [13]FISER J, BERKES P, WHITE B. No evidence for active sparsification in the visual cortex[C] / / Advances in Neu⁃ ral Information Processing Systems, Vancouver, B. C., Canada, 2009: 108⁃116. [14]ZHANG D, YANG M, FENG X. Sparse representation or collaborative representation: which helps face recognition? [C] / / Proceedings of the IEEE International Conference on Computer Vision. Barcelona, Spain, 2011, 6669(5): 471⁃478. [15] DENG W, HU J, GUO J. Extended SRC: undersampled face recognition via intraclass variant dictionary[ J]. Pat⁃ tern analysis and machine intelligence, IEEE transactions on, 2012, 34(9): 1864⁃1870. [16]GUO J. In defense of sparsity based face recognition[C] / / Proceedings of the IEEE International Conference on Com⁃ puter Vision. Sydney, NSW, Australia, 2013, 9(4):399⁃ 406. [17]SU Y, SHAN S, CHEN X, et al. Adaptive generic learn⁃ ing for face recognition from a single sample per person [C] / / Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition. San Francis⁃ co, USA, 2010: 2699⁃2706. [18] WEI C, WANG Y⁃F. Learning auxiliary dictionaries for undersampled face recognition [ C] / / Proceedings of the IEEE International Conference on Multimedia and Expo. San Jose, California, USA, 2013, 2013:1⁃6. [19] WANG X, TANG X. Unified subspace analysis for face recognition[C] / / null. Proceedings of the IEEE Interna⁃ tional Conference on Computer Vision. Nice, France, 2003:679⁃686. [20] LECUN Y L, BOTTOU L, BENGIO Y, et al. Gradient⁃ based learning applied to document recognition. Proc IEEE 第 3 期 马晓,等:基于深度学习特征的稀疏表示的人脸识别方法 ·285·
<<向上翻页向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有