摘要
面部表情是一种重要的肢体语言,在人们的日常生活中,只有7%的信息是通过语言来传递的,而55%的信息则是通过面部表情来传递的。人脸表情识别就是利用计算机提取人脸表情图像的特征信息,根据特征的不同将表情图像归属到7种不同的表情类别中,它使得计算机能够根据表情图像分类的结果,推断人的心理状态,从而实现人机之间的自然交互。尽管目前人脸表情识别技术已经取得了不少进展,但现实生活中光照、姿态、噪声、遮掩物等各种因素影响,要实现大规模的应用仍需进一步研究。
本文分析了人脸表情识别技术国内外研究现状,对计算机人脸表情识别的若干问题进行了探讨,着重研究了流形学习方法在人脸表情识别中的应用,并进行了一系列表情识别实验。论文的研究工作主要包括以下几个方面:
1.对流形学习非线性降维方法作了重点介绍,详细介绍了几种常用的流形学习方法,如:等距映射、局部线性嵌入、拉普拉斯特征映射、海赛局部线性嵌入和局部切空间排列,并分析了其主要优点与不足。
2.提出了基于Contourlet变换与局部线性嵌入的人脸表情识别方法。对Contourlet变换理论作了详细介绍,并将其应用于人脸表情特征提取,生成具有多分辨率、多尺度的人脸表情特征。采用LLE算法进行特征降维,在JAFFE数据库和Cohn-Kanada数据库上进行了实验,与Wavelet+LLE+SVM和PCA+SVM进行了比较,本文提出的Contourlet+ LLE的非特定人的人脸表情识别方法在JAFFE数据库和Cohn-Kanada数据库上的最高识别率分别可以达到63.81%和69.1%,均高于以上两种方法。
3.分别对原始LBP算子、多分辨率LBP算子、旋转不变LBP算子和均匀模式LBP算子作了介绍,分析了各自的优点与不足。重点叙述了均匀模式LBP算子的人脸表情图像特征提取应用。
4.提出了基于局部二元模式的拉普拉斯特征映射人脸表情识别方法。介绍了一种称为“图嵌入”的数据降维框架法,通过设定特定的近邻图矩阵来重新构造LE算法。在JAFFE数据库与Cohn-Kanada数据库上进行了大量非特定人的人脸表情识别实验,分析了LBP算子参数(P,R)和LBP特征图像方格划分方式对实验结果的影响,将LE算法与线性降维方法PCA和LDA进行了比较,本文提出的LBP+LE算法在JAFFE数据库和Cohn-Kanada数据库上的识别率分别可达到70.48%和70.95%,均高于LBP+PCA和LBP+LDA的最高识别率,验证了LE算法的有效性。
Facial expression is an important body language. In our daily life, only 7 percent of information is expressed by oral language and 55 percent of information is expressed by facial expression. The major work of facial expression recognition is using computer to extract facial' features of all expression images. Then acrroding to the difference of features, each image is classified to one class of seven different expressions. It makes computer know the expression states from the classify result and achieve Human-Computer Interaction. Much progress has been got, but, in real life, on the influence of _illumination_, posture, noise, masking, and so on, facial expression recognition technology still needs to do more researches to achieve praticail applications.
In this paper,we analyzed the process of domestic and international facial expression recognition technology in recent years and discussed several problems about facial expression recognition by computer. The application of Manifold learning method in facial expression recognition is detailedly introduced and a series of experiments are carried out. The research work in this paper mainly includes the following several respects:
1. A nonlinear dimensional reduction method named Manifold Learning was particularly introduced. Some of classical Manifold learning algorithms, such as Isomap, locally linear embedding, Laplacian eigenmaps, Hessian-based locally linear embedding and local tangent space alignment, were recommended in detail. The advantages and disadvantages of above Manifold Learning algorithms was analysised.
2. A facial expression recognition method based on Contourlet Transform and Locally Linear Embedding was prresented. It explicitly introduced the theory of Contourlet Transform. Using Contourlet Transform for facial expression feature extraction, it generated the multiresolution and multiscale feature of original image. LLE algorithm was used for feature dimensional reduction. Experiments were carried out on JAFFE database and Cohn-Kanada database. Compared with Wavelet+LLE+SVM and PCA+SVM, the maximum recognition rates for facial expression recognition of non-given person of CT+LLE on JAFFE database and Cohn-Kanada database respectively are 63.81 percent and 69.1 percent, which is higher than that of two methods.
3. Original Local Binary Pattern (LBP) operator, multiresolution LBP operator, rotation invariance LBP operator and uniform LBP operator were introduced and the advantages and disadvantages of them was analyzed. We mainly introduced the application of uniform LBP operator for facial expression feature extraction.
4. A facial expression recognition method with Local Binary Pattern and Laplacian Eigenmaps was presented. It introduced a framework algorithm for data dimensional reduction which named graph embedding. The Laplacian Eigenmaps algorithm was recomposed by designing the neighboring weight matrix. Plenty of experiments for non-given person facial expression recognition were carried out on JAFFE database and Cohn-Kanada database. The influences of LBP parameters (P,R) and block dividing on experimental result was analyzed. Compared LE with PCA and LDA, the maximum recognition rates of LBP+LE on JAFFE database and Cohn-Kanada database respectively are 70.48 percent and 70.95 percent, which are both higher than LBP+PCA and LBP+LDA. It proves the method presented in this paper is effective and feasible.
引文
[1]Albert M. Communication without Words [J]. Psychology Today,1968,2(4):53-56.
[2]Haisong Gu, Qiang Ji and Zhiwei Zhu. Active Facial Tracking for Fatigue Detection [A]. In Proc.6th IEEE WACV [C],2002, Orlando, Florida, USA.
[3]刘晓旻,谭华春,章毓晋.人脸表情识别研究的新进展[J].中国图象图形学报,2006,11(10):1360-1367.
[4]何良华,邹采荣,包永强等.人脸面部表情识别的研究进展[J].电路与系统学报,2005,10(1):70-73.
[5]Darwin C. The Expression of the Emotions in Man and Animals [M]. J. Murray, London,1872.
[6]Ekamn P, Friesen W V. Facial Action Coding System (FACS):manual [M]. Consulting Psychologists Press,1978.
[7]宋星光.人脸表情识别研究[D].长沙:中南大学,2005.
[8]吴丹,林学阖.人脸表情视频数据库的设计与实现[J].计算机工程与应用,2004,5:177-180.
[9]张庆凯.人脸表情视频数据库系统的实现及相关算法研究[D].沈阳:东北大学,2005.
[10]Essa Irfan A. Coding, Analysis, Interpretation, and recognition of Facial Expressions [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1997-07,19(7):757-763.
[11]周书仁,梁昔明,杨秋芬等.类间学习神经网络的人脸表情识别[J].计算机应用研究,2008,25(7):2219-2223.
[12]Turk M, Pentland A. Face recognition using eigenfaces [A]. In:Proc Computer Vision and Pattern Reorganization Conferance [C],1991,586-591.
[13]Cootes T, Taylor C, Cooper D, Graham J. Active shape models-their training and applications [J]. Computer Vision and Image Understanding,1995,61(1):637-644.
[14]Fasel B, Luettin J. Automatic facial expression analysis:A survey [J], Pattern Recognition Society,2003,36:259-275.
[15]Yin L, Basu A. Integrating active face tracking with model based coding [J]. Pattern Recognition Letter,1999,20:651-657.
[16]Lee C H, Kim J S, Park K H. Automatic human face location in a complex background [J]. Pattern Recognition,1996,29:1877-1889.
[17]Hok-chun Lo, Ronald Chung. Facial expression recognition approach for performance animation [J]. In:Proc. The Second Internet Workshop on Digital and Computational Video,2001.pp:132-139.
[18]Andrew A, Calder J, Burton M. A Principal Component Analysis of Facial Expressions [J]. Vision Research,2001,41:1179-1208.
[19]Havran C, et al. Independent Component Analysis for face authentication [A]. KES'2002 Proceedings-Knowledge-Based Intelligent Information and Engineering Systems[C]. Crema (Italy),2002(09):1207-1211.
[20]Cootes T F,Edwards G J. Active appearance models [A].In:5th European Conference on Computer Vision[C].1998:484-498.
[21]Huang C L and Huang Y M. Facial Expression Recognition Using Model-Based Feature Extractio n and Action Parameters Classification [J]. Visual Comm. and Image Representation,1997,8(3):278-290.
[22]Queen C, Thomas H. Facial expression recognition:A clustering-based approach Pattern [J]. Recognition Letters,2003,24:1295-1302.
[23]Essa I A. Coding, Analysis, Interpretation, and recognition of Facial Expressions [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1997-07, 19(7):757-763.
[24]金辉.人脸面部表情编码、分析、识别的研究与实现[D].哈尔滨工业大学博士学位论文,2000:25-30.
[25]Fabio L, Roberto P. An Efficient Use of MPGE-4 FAP Interpolation for Facial Animation at 70 bits/Frame [J]. IEEE Transactions on Circuits and Systems for Video Technology,2001,11(10).
[26]姚伟,孙正兴,张岩.面向脸部表情识别的Gabor特征选择方法[J].计算机辅助设计与图形学学报,2008,20(1):79-84.
[27]任金霞,杨国亮.基于Gabor和ADABOOST的面部表情识别[J].微计算机信息,2007,23(3-1):290-292.
[28]姚雪梅.基于小波变换的局部PCA人脸识别研究[D].新疆:新疆大学,2006.
[29]Duthoit C J, Sztynda T, Lal S K L, Jap B T, Agbinya J I.Optical flow image analysis of facial expressions of human emotion:forensic applications. [A]. Proceedings of the 1st international conference on Forensic applications and techniques in telecommunications, information, and multimedia and workshop[C]. ICST.2008,5.
[30]余棉水,黎绍发.基于光流的动态人脸表情识别[J].微电子学与计算机,2005,22(7):113-119.
[31]Lien J J, Takeo Kanade.Automated facial expression recognition based on FACS action units [A]. In:Proceedings of the 3rd International Conference on Automatic Face and Gesture Recognition [C].1998,390-395.
[32]尹星云,王询,董兰芳.用隐马尔可夫模型设计人脸表情识别系统[J].电子科技大学学报,2003,32(6):725-728.
[33]Ira C, Ashutosh G, Thomas S H. Emotion Recognition from Facial Expressions using Multilevel HMM [J]. Beckman Institute for Advanced Science and Technology, The University of Illin oisat Urbana-Champaign.2000,1-7.
[34]Reddy K R L, Babu G R, Kishore L, Maanasa M..MULTISCALE FEATURE AND SINGLE NEURAL NETWORK BASED FACE RECOGNITION. [J]. Journal of Theoretical and Applied Information Technology,2005-2008,571-577.
[35]Sun T H, Tien F C..Using backpropagation neural network for face recognition with 2D+3D hybrid information. [J]..Expert Systems with Applications,2008,35(1-2): 361-372.
[36]凌旭峰,杨杰,叶晨洲.基于支撑向量机的人脸识别技术[J].红外与激光工程,2001,30(5):319-327.
[37]Tenenbaum J B, Silva V, Langford J C. A global geometric framework for nonlinear dimensionality reduction [J]. Science,2000,290(5500):2319-2323.
[38]Seung H S, Lee D D. The manifold ways of perception [J]. Science,2000,290 (5500):2268-2269.
[39]Rowies S T, Lawrance K S. Nonlinear Dimensionality reduction by locally linear embedding [J]. Science,2000,290(5500):2323-2326.
[40]Donoho D L, Grimes C. Hessian eigenmaps:New locally linear embedding techniques for high dimensional data [J]. Proeeedings of the National Aeademy of Seienees,2003,100(10):5591-5596.
[41]Belkin M, Niyogi P. Laplacian Eigenmaps for Dimensionality Reduction and Data Representation [J]. Neural Computation,2003,51(6):1373-1396.
[42]Zhang Z, Zha H. Principal Manifolds and Nonlinear Dimension Reduetion via Local Tangent Space Alignment [J]. SIAM Journal of Seientifie Computing,2004, 26(1):313-338.
[43]朱明旱,罗大庸.基于流形的表情分解算法[J].计算机工程与应用,2008,44(32):203-205.
[44]张军平,曹存根.神经网络及其应用[M].北京:清华大学出版社,2004.
[45]Williams C. On a connection between kernel PCA and metric multidimensional scaling [J]. Machine Learing,46:11-19,2002.
[46]陈省身,陈维桓.微分几何讲义[M].北京:北京大学出版社,1983.
[47]Silva V, Tenenbaum J. Globa Iversus Local Methods in Nonlinear Dimensionality Reduetion [J]. Advances in Neural Information Processing Systems, 2003,15:705-712.
[48]Ham J, Lee D D, Mika S and Scholkopf B. A kernel view of the dimensionality reduction of manifolds [J]. Technical Report TR-110, Max-Planck-Institut fur biologische Kybernetik, Tubingen, July 2003.
[49]王靖.流形学习的理论与方法研究[D].杭州:浙江大学,2006.
[50]Lawrence K S, Sam T R. Think Globally, Fit Locally:Unsupervised Learning of Low Dimensional Manifolds [J]. Journal of Machine Learning Research,2003, 4:119-155.
[51]Belkin M and Niyogi P. Laplacian Eigenmaps and Spectral Techniques for Embedding and Clustering [J]. Neuarl Information Porcessings Systems 14 (NIPS'2001), PP.585-591,2002.
[52]Guo Y, Cao J B, Kwan W H P. Kernel Laplacian eigenmaps for visualization of non-vectorial data [C]. Proceedings of Conference on Artificial Intelligence, 2006:1179-1183.
[53]Minh N. Do, Vetterli M. The Contourlet Transform:An Efficient Directional Multiresolution Image Representation [J]. IEEE Transactions on Image Processing, 2005,14(12):2091-2106.
[54]Minh N.Do, Vetterli M. Contourlets:a directional multiresolution image representation [J]. International Conference on Image Processing, Beckman Inst., Illinois Univ., Urbana, IL, USA, December,2002:357-360.
[55]Kun Liu, Lei Zhang and Weiwei Chang. Regional Feature Self-Adaptive Image Fusion Algorithm Based on Contourlet Transform [J]. ACTA OPTICA SINICA, 2008,28(4):681-686.
[56]Minh N.Do, Vetterli M. Pyramidal directional filter banks and curvelets [J]. In Proc. IEEE Int. Conf. on Image Proc., Thessaloniki, Greece, Oct,2001.
[57]Bamberger R H, Smith M J T. A filter bank for the directional decomposition of images:Theory and design [J]. IEEE Trans. Signal Proc.,1992,40(4):882-893.
[58]Burt P, Adelson E H. The Laplacian pyramid as a compact image code [J]. IEEE Trans. Commun.,1983,31(4):532-540.
[59]Do M N and Vetterli M. Framing pyramids [J]. IEEE Trans. Signal Proc.,2003,9: 2329-2342.
[60]Heureux P J L, Carreau J, Bengio Y, et.al. Locally Linear Embedding for dimensionality reduction in QSAR[J]. Journal of Computer-Aided Molecular Design,2004,18(5):475-482.
[61]Junping Zhang. Manifold Learning and Applications [D]. PhD Thesis, Institute of Automation, Chinese Academey of Sciences, Beijing, China,2003.
[62]Sergios Theodoridis and Konstantinos Koutrounmbas, Pattern Recognition, Third Edition [M], Publishing House of Electronics Industry, Beijing,2006.
[63]Yang G L, Wang Z L, Wang G J. A Survey of Facial Expression Recognition [J]. Techniques of automation and applications.2006:25(4):1-6.
[64]Yan S C, Xu D, Zhang B Y, et al. Graph Embedding:A General Framework for Dimensionality Reduction [J]. In:IEEE Computer Society Conference 2005 on Computer Vision and Pattern Recognition.2005,2:830-837.
[65]Ahonen T, Hadid A, Poetikainen M. Face description with local binary patterns: Application to face recognition [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2006,28(12):2037-2041.
[66]Ojala T, Pietikainen M, Maenpaa T. Multiresolution Gray Scale and Rotation Invariant Texture Classification with Local Binary Pattern [J]. Pattern. Analysis and Machine Intelligence, IEEE Transaction.2002,24(7):971-987.
[67]Matti Pietikainen. The local binary pattern approach to texture analysis-extensions and applications [M]. University of Oulu:2003.
[68]Ahonen T, Hadid A, Pietikainen M. Face Recognition with Local Binary Patterns [J]. LNCV, ECCV 2004:469-481.