基于FKPCA+双子空间和信息属性KNN的小样本人脸识别研究
|
摘要
人脸识别技术因其具有重要的科学意义和实用价值,在近几年得到了研究者的高度重视,成为当前模式识别和人工智能领域的一个研究热点。人脸识别主要分为人脸检测、特征提取和模式分类三个步骤。人脸识别过程中会遇到各种问题,其中样本维数过高、类别数大、单人训练样本少以及识别的实时性都是亟待解决的难题。
本文在综合分析了以往的人脸识别方法的基础上,着重从特征提取速度、特征信息的完备性和识别率三个方面考虑,提出基于FKPCA+双子空间和信息属性KNN分类器的小样本人脸识别算法,并设计实现相应的原型系统。本文的工作主要包括:
(1)提出一种FKPCA+双子空间人脸特征信息提取新方法。该方法首先通过快速核主元分析(FKPCA)将低维空间线性不可分的人脸样本隐性映射到线性可分的高维空间,并在该空间中实现降维;然后,在FKPCA降维后的数据上,利用Fisher准则从值域空间中提取常规特征信息,利用类间散度准则从零空间提取非常规特征信息。该方法可提取到更加完备的特征信息,对提升识别率有很大的帮助。
(2)提出信息属性K-近邻分类器算法。在深入研究传统的K-近邻分类器算法的基础上,针对K-近邻分类器中欧式距离不能很好的表示样本间相似度问题,采用以信息属性为权值对欧式距离进行加权的策略,有效地改进欧式距离;把加权后的欧式距离作为K-近邻分器度量样本间相似度的测度。该方法能够有效地弥补传统欧氏距离仅仅表示m维空间中两个点之间真实距离而不能很好地表示对象间相似程度的不足。
(3)提出一种小样本人脸识别算法。首先使用FKPCA对样本进行线性降维;然后提取降维后数据的常规特征信息和非常规特征信息;接着利用改进后的欧式距离对常规特征信息和非常规特征信息分别计算测试样本到各训练样本的相似度,并把计算结果进行融合;最后利用K-近邻分类器使用融合结果进行分类。
(4)人脸识别原型系统的实现。采用面向对象思想设计并开发了小样本人脸识别原型系统。该系统由图像预处理、FKPCA特征提取、双子空间两类信息提取和分类识别四个功能模块组成,实现了当训练样本较少时,系统仍能保持较高的正确识别率和较好的实时性。
Face recognition technology has been attached great importance to the researchers for its Scientific significance and practical value in the past few years,and become the hotspot of current pattern recognition and artificial intelligence.Face recognition normally be regarded as have three processes that are face detection, features extraction and pattern classification.Face recognition often meet these problems,the dimension of sample too high,the classes of pattern too mach and each person could only provide a small amount training sample.
In this paper,by systematically analyzing of relevant algorithms,we present some novel algorithms for face recognition with small sample based on FKPCA+Double Subspace and Information Attribute KNN classifier,from three aspects of the speed of extracting features,information completeness and face recognition.In addition,a prototype system of face recognition is designed and implemented.The highlights and main contributions of the dissertation include:
(1) A novel method based on FKPCA+Double Subspace for features extraction is presented.Firstly,FKPCA is used to map input space information to High-dimensional space to reduce the dimension of original samples in High-dimensional space;secondly,get the regular information using Fisher Criterion in Rang Space and gain the Irregular information employing between-class scatter Criterion in Null Space.This method can extract more complete optimal discriminant features,and be also of great help to the feature extraction problem in small sample case.
(2) The Information Attribute K-nearest Neighbor classifier is studied.On the base of thorough study of traditional K-nearest neighbor classifier,against the problem of Euclidean Distance of KNN can not exptress sample semblable degree well,the Strategy of using information attribute as Weights is adopted to improve Euclidean Distance;then the impoved Euclidean Distance is used as K-nearest neighbor classifier measure.The Information Attribute K-nearest Neighbor classifier can effectively make up the fault that Euclidean Distance render only the truly distance of two points in m dimension space.
(3) An algorithm on Face Recognition with Small Sample is given. Firstly,FKPCA is used to implement linearly Dimension Reduction;secongly, extract Regular information and Irregular information;thridly,Information Attribute Euclidean Distance is used to carry out of calculating the sample semblable degree with Regular information and Irregular information alone.Then,the calculations have Worked out are fused.At last,K-nearest neighbor classifier the task of classification with the fused result.
(4) Based on the idea of object-oriented,we design and development a prototype system of face recognition with small Sample,which is divided into four modules which are image preprocess,FKPCA process,Double Subspace feature extraction and Information Attribute KNN face recognition.And makes the system recognize people according to face image with only a little training samples.System has managed to maintain a higher correct recognition rate.
本文在综合分析了以往的人脸识别方法的基础上,着重从特征提取速度、特征信息的完备性和识别率三个方面考虑,提出基于FKPCA+双子空间和信息属性KNN分类器的小样本人脸识别算法,并设计实现相应的原型系统。本文的工作主要包括:
(1)提出一种FKPCA+双子空间人脸特征信息提取新方法。该方法首先通过快速核主元分析(FKPCA)将低维空间线性不可分的人脸样本隐性映射到线性可分的高维空间,并在该空间中实现降维;然后,在FKPCA降维后的数据上,利用Fisher准则从值域空间中提取常规特征信息,利用类间散度准则从零空间提取非常规特征信息。该方法可提取到更加完备的特征信息,对提升识别率有很大的帮助。
(2)提出信息属性K-近邻分类器算法。在深入研究传统的K-近邻分类器算法的基础上,针对K-近邻分类器中欧式距离不能很好的表示样本间相似度问题,采用以信息属性为权值对欧式距离进行加权的策略,有效地改进欧式距离;把加权后的欧式距离作为K-近邻分器度量样本间相似度的测度。该方法能够有效地弥补传统欧氏距离仅仅表示m维空间中两个点之间真实距离而不能很好地表示对象间相似程度的不足。
(3)提出一种小样本人脸识别算法。首先使用FKPCA对样本进行线性降维;然后提取降维后数据的常规特征信息和非常规特征信息;接着利用改进后的欧式距离对常规特征信息和非常规特征信息分别计算测试样本到各训练样本的相似度,并把计算结果进行融合;最后利用K-近邻分类器使用融合结果进行分类。
(4)人脸识别原型系统的实现。采用面向对象思想设计并开发了小样本人脸识别原型系统。该系统由图像预处理、FKPCA特征提取、双子空间两类信息提取和分类识别四个功能模块组成,实现了当训练样本较少时,系统仍能保持较高的正确识别率和较好的实时性。
Face recognition technology has been attached great importance to the researchers for its Scientific significance and practical value in the past few years,and become the hotspot of current pattern recognition and artificial intelligence.Face recognition normally be regarded as have three processes that are face detection, features extraction and pattern classification.Face recognition often meet these problems,the dimension of sample too high,the classes of pattern too mach and each person could only provide a small amount training sample.
In this paper,by systematically analyzing of relevant algorithms,we present some novel algorithms for face recognition with small sample based on FKPCA+Double Subspace and Information Attribute KNN classifier,from three aspects of the speed of extracting features,information completeness and face recognition.In addition,a prototype system of face recognition is designed and implemented.The highlights and main contributions of the dissertation include:
(1) A novel method based on FKPCA+Double Subspace for features extraction is presented.Firstly,FKPCA is used to map input space information to High-dimensional space to reduce the dimension of original samples in High-dimensional space;secondly,get the regular information using Fisher Criterion in Rang Space and gain the Irregular information employing between-class scatter Criterion in Null Space.This method can extract more complete optimal discriminant features,and be also of great help to the feature extraction problem in small sample case.
(2) The Information Attribute K-nearest Neighbor classifier is studied.On the base of thorough study of traditional K-nearest neighbor classifier,against the problem of Euclidean Distance of KNN can not exptress sample semblable degree well,the Strategy of using information attribute as Weights is adopted to improve Euclidean Distance;then the impoved Euclidean Distance is used as K-nearest neighbor classifier measure.The Information Attribute K-nearest Neighbor classifier can effectively make up the fault that Euclidean Distance render only the truly distance of two points in m dimension space.
(3) An algorithm on Face Recognition with Small Sample is given. Firstly,FKPCA is used to implement linearly Dimension Reduction;secongly, extract Regular information and Irregular information;thridly,Information Attribute Euclidean Distance is used to carry out of calculating the sample semblable degree with Regular information and Irregular information alone.Then,the calculations have Worked out are fused.At last,K-nearest neighbor classifier the task of classification with the fused result.
(4) Based on the idea of object-oriented,we design and development a prototype system of face recognition with small Sample,which is divided into four modules which are image preprocess,FKPCA process,Double Subspace feature extraction and Information Attribute KNN face recognition.And makes the system recognize people according to face image with only a little training samples.System has managed to maintain a higher correct recognition rate.
引文
[1]Wechsler H,Philips P J,Bruce Vetal.Face recognition from theory to application.Berlin:Springer Verlag,1998,2(11):402-411.
[2]Samal A,lyengar P A.Automatic recognition and analysis for human faces and facial expressions.Pattern Recognition,1992,25(8):65-77
[3]Valentin D.Connectionist models of face processing.Pattern Recognition,1994,27(8):1209-1230
[4]陈刚,戚飞虎.实用人脸识别系统的本征脸法实现.北京:计算机研究与发展,2001,38(2):170-174
[5]周激流,张晔.人脸识别理论研究进展.北京:计算机辅助设计与图形学学报,1999,11(2):180-184
[6]李刚,高政.人脸自动识别方法综述.四川:计算机应用研究,2003,8(2):4-9
[7]Yang Y Z,Huang T S.Human face detection in a complex background.Pattern Recognition,1994,27(1):705-763
[8]A.Lanitis,C.J.Taylor and T.F.Cootes.Automatic Identification and Coding of Human Faces using Flexible Models.IEEE Transactions on Pattern Analysis and Machine Intelligence.1997,19(7):743-756.
[9]Andress Lanitis,Chris J Taylor,Automatic interpretation and coding of face images using flexible models.IEEE Trans on PAMI,1997,19(7):743-766
[10]J.Wu,Z-H.Zhou.Face Recognition with one training image per person.Pattern Recognition Letters,23(14),1711-1719,2002.
[11]S.Chen,D.Zhang,and Z.-H.Zhou,Enhanced(PC)~2 A for face recognition with one training image per person,Pattern Recognit.Lett.,vol.25,pp.1173-1181,2004.
[12]Chen,S.,Zhang,D.and Zhou,Z.-H.:A new face recognition methodbased on SVD perturbation for single example image per person,Appl.Math.Comput.2005.
[13]Peter N.Belhumeur,Joao P.Hespanha,David J.Kriegman.Eigenfaces vs.Fisherfaces:Recognition using class specific linear projection.IEEE Trans.PatternAnal.Machine Intell.19(7)(1997) 711-720
[14]R.A.Fisher,The use of multiple measurements in taxonomic problems,Annals of Eugenics 7(1936) 178-188.
[15]Foley D H and Sammon J W Jr.,An optimal set of discriminant vectors,IEEE Trans.24(3)(1975)281-289.
[16]J.W.Sammon.An optimal discriminant plane,IEEE Trans.Computer,C-19(1970) 826-829.
[17]金忠,杨静宇,陆建峰.一种具有统计不相关性的最优鉴别矢量集.计算机学报,1999,22(10):1105-1108
[18]Zhong Jin,J.Y Yang,Z.S.Hu,Z.Lou,Face Recognition based on uncorrelated discriminant transformation,Pattern Recognition,2001,34(7),1405-1416
[19]Z.Jin,J.Y Yang,Z.M.Tang,Z.S.Hu,A theorem on uncorrelated optimal Discriminant vectors,Pattern Recognition,2001,34(10),2041-2047.
[21]Scholkopf,B.,Smola,A.Learning with Kernels.MIT.2002.
[22]G.Guo,S.Z.Li,Kap Luk Chan.Support vector machines for face recognition.Image and Vision Computing.19.2001:631-638
[23]G.Baudat,F.Anouar,Generalized discriminant analysis using a kernel approach,Neral Computation,2000,12(10):2385-2404.
[24]Ming-Hsuan Yang.Kernel Eigenfaces vs.Kernel Fisherfaces:Face Recognition Using Kernel Methods.Proceedings of the fifth IEEE International Conference on Automatic Face and gesture recognition.2002.215-220
[25]Jian Yang,Alejandro F.Frangi,Jing-yu Yang et al,KPCA Plus LDA:A Complete Kernel Fisher Discriminant Framework for Feature Extraction and Recognition.IEEE Trans.Pattern Anal.Machine Intell.Vol.27,No.2,2005,230-244.
[26]高秀梅,杨静宁,金忠,陈才扣.基于核的Foley-Sammon鉴别分析与人脸识别.计算机辅助设计与图形学学报,2004,16(7):962-967
[27]Yong Xu,Jing-yu Yang,Jianfeng Lu,Dong-jun Yu.An efficient renovation on kernel Fisherdiscriminant analysis and face recognition experiments.Pattern Recognition,37,2091-2094,2004
[28]Buhmann J et al.Size and distortion invariant object recognition by hierarchical graph matching,Intel.Conf,on Neural Networks,1990,411-416.
[29]M.Lades,et al,Distortion invariant object recognition in the dynamic link architecture,IEEE Trans.Comput.,vol.42,pp.300-311.
[30]Kohonen T.Self-organization and associative memory.Berlin:Springer,1988
[31]F.S.Samaria.Face recognition using hidden Markov models,PHD thesis.Trinity College,University of Cambridge,Cambridge,1994
[32]刘小军.人脸识别技术研究[博士学位论文].中国科学院电子学研究所,2001-05-18
[33]杨竹青,李勇,胡德文.独立成分分析方法综述.自动化学报,2002,28(5):762-772
[34]Bartlett M S,Movellan J R,Sejnowski T J.Face Recognition by Independent Component Analysis.IEEE Trans.on Neural Network,2002 13(6):1450-1464.
[35]Bartlett M,Lades H,Sejnowski T.Independent component representations for face recognition.ln:Proceedings of the SPIE Symposium on Electronic Imaging:Human Vision and Electronic Imaging.SanJose,California,1998,3299-3310
[36]A.H.Gunatilaka and B.A.Baertlein,Feature-level and decision-level fusion of noncoincidently sampled sensors for land mine detection,IEEE Trans.PAMI,23(6)2001.577-589
[37]Baudat G,Anouar F.Kernel-based methods and function approximation[C].Proc of the International Joint Conf on Neural Networks.Piscataway:Institute of Electrical and Electronics Engineers Inc,2001:1244-1249..
[38]廖频.基于统一概率模型的人脸识别技术[博士学位论文].中国科学院研究生院博士论文,2003.
[39]张敏贵,潘泉,张洪才等.多生物特征识别[J].信息与控制,2002,31(6):524-528.
[40]http://www biometriegroup.Eom
[41]陈振洲,李磊,姚正安.基于SVM的特征加权KNN算法[J].中山大学学报(自然科学版),2005,(01)
[42]COVER T M,HART P E.Nearest neighbor pattern classification[J].In Trans IEEE InformTheory,1967,IT-13:21-27.
[43]K.Etemad and T.Chellappa,"Discriminant Analysis for Recognition of Human Face Images,"Journal of the Optical Society of America A,Vol.14,pp.1724-1773,1997.
[44]Klema V C.The Singular Value Decomposition:Its Computation and Some Applications[J].IEEE Transactions on Automatic Control,1980,25(2):164-176.
[45]K.Etemad and T.Chellappa,"Discriminant Analysis for Recognition of Human Face Images,"Journal of the Optical Society of America A,Vol.14,pp.1724-1773,1997.
[46]Li Ming,Yuan Baozong.2D-LDA:A statistical linear discriminant analysis for image matrix.Pattern Recognition Letters,26(2005)5,527-5
[47]Yang Jian,Zhang David,Yong Xu,Yang Jingyu.Two-dimensional discriminant transform for face recognition.Pattern Recognition,38(2005)7,1125-1129.
[48]Kimura F,Shridhar M.Handwritten number recognition based on multiple algorithms.Pattern Recognition,1991,24(10):969-983
[49]A.Nefian,M.H.Hayes Ⅲ.An Embedded HMM-Based Approach for Face Detection and Recognition.Proceedings of IEEE International Conference on Acoustics,Speech,and Signal Processing.1999,6(6):3553-3556.
[50](英)[L.拉宾纳]Lawrence Rabiner,(英)[阮平望]Biing-Hwang Juang著.语音识别基本原理.北京:清华大学出版社,1999.
[51]Samaria F.Face Recognition Using Hidden Markov Models.Engineering Department,Cambridge University.1994.
[2]Samal A,lyengar P A.Automatic recognition and analysis for human faces and facial expressions.Pattern Recognition,1992,25(8):65-77
[3]Valentin D.Connectionist models of face processing.Pattern Recognition,1994,27(8):1209-1230
[4]陈刚,戚飞虎.实用人脸识别系统的本征脸法实现.北京:计算机研究与发展,2001,38(2):170-174
[5]周激流,张晔.人脸识别理论研究进展.北京:计算机辅助设计与图形学学报,1999,11(2):180-184
[6]李刚,高政.人脸自动识别方法综述.四川:计算机应用研究,2003,8(2):4-9
[7]Yang Y Z,Huang T S.Human face detection in a complex background.Pattern Recognition,1994,27(1):705-763
[8]A.Lanitis,C.J.Taylor and T.F.Cootes.Automatic Identification and Coding of Human Faces using Flexible Models.IEEE Transactions on Pattern Analysis and Machine Intelligence.1997,19(7):743-756.
[9]Andress Lanitis,Chris J Taylor,Automatic interpretation and coding of face images using flexible models.IEEE Trans on PAMI,1997,19(7):743-766
[10]J.Wu,Z-H.Zhou.Face Recognition with one training image per person.Pattern Recognition Letters,23(14),1711-1719,2002.
[11]S.Chen,D.Zhang,and Z.-H.Zhou,Enhanced(PC)~2 A for face recognition with one training image per person,Pattern Recognit.Lett.,vol.25,pp.1173-1181,2004.
[12]Chen,S.,Zhang,D.and Zhou,Z.-H.:A new face recognition methodbased on SVD perturbation for single example image per person,Appl.Math.Comput.2005.
[13]Peter N.Belhumeur,Joao P.Hespanha,David J.Kriegman.Eigenfaces vs.Fisherfaces:Recognition using class specific linear projection.IEEE Trans.PatternAnal.Machine Intell.19(7)(1997) 711-720
[14]R.A.Fisher,The use of multiple measurements in taxonomic problems,Annals of Eugenics 7(1936) 178-188.
[15]Foley D H and Sammon J W Jr.,An optimal set of discriminant vectors,IEEE Trans.24(3)(1975)281-289.
[16]J.W.Sammon.An optimal discriminant plane,IEEE Trans.Computer,C-19(1970) 826-829.
[17]金忠,杨静宇,陆建峰.一种具有统计不相关性的最优鉴别矢量集.计算机学报,1999,22(10):1105-1108
[18]Zhong Jin,J.Y Yang,Z.S.Hu,Z.Lou,Face Recognition based on uncorrelated discriminant transformation,Pattern Recognition,2001,34(7),1405-1416
[19]Z.Jin,J.Y Yang,Z.M.Tang,Z.S.Hu,A theorem on uncorrelated optimal Discriminant vectors,Pattern Recognition,2001,34(10),2041-2047.
[21]Scholkopf,B.,Smola,A.Learning with Kernels.MIT.2002.
[22]G.Guo,S.Z.Li,Kap Luk Chan.Support vector machines for face recognition.Image and Vision Computing.19.2001:631-638
[23]G.Baudat,F.Anouar,Generalized discriminant analysis using a kernel approach,Neral Computation,2000,12(10):2385-2404.
[24]Ming-Hsuan Yang.Kernel Eigenfaces vs.Kernel Fisherfaces:Face Recognition Using Kernel Methods.Proceedings of the fifth IEEE International Conference on Automatic Face and gesture recognition.2002.215-220
[25]Jian Yang,Alejandro F.Frangi,Jing-yu Yang et al,KPCA Plus LDA:A Complete Kernel Fisher Discriminant Framework for Feature Extraction and Recognition.IEEE Trans.Pattern Anal.Machine Intell.Vol.27,No.2,2005,230-244.
[26]高秀梅,杨静宁,金忠,陈才扣.基于核的Foley-Sammon鉴别分析与人脸识别.计算机辅助设计与图形学学报,2004,16(7):962-967
[27]Yong Xu,Jing-yu Yang,Jianfeng Lu,Dong-jun Yu.An efficient renovation on kernel Fisherdiscriminant analysis and face recognition experiments.Pattern Recognition,37,2091-2094,2004
[28]Buhmann J et al.Size and distortion invariant object recognition by hierarchical graph matching,Intel.Conf,on Neural Networks,1990,411-416.
[29]M.Lades,et al,Distortion invariant object recognition in the dynamic link architecture,IEEE Trans.Comput.,vol.42,pp.300-311.
[30]Kohonen T.Self-organization and associative memory.Berlin:Springer,1988
[31]F.S.Samaria.Face recognition using hidden Markov models,PHD thesis.Trinity College,University of Cambridge,Cambridge,1994
[32]刘小军.人脸识别技术研究[博士学位论文].中国科学院电子学研究所,2001-05-18
[33]杨竹青,李勇,胡德文.独立成分分析方法综述.自动化学报,2002,28(5):762-772
[34]Bartlett M S,Movellan J R,Sejnowski T J.Face Recognition by Independent Component Analysis.IEEE Trans.on Neural Network,2002 13(6):1450-1464.
[35]Bartlett M,Lades H,Sejnowski T.Independent component representations for face recognition.ln:Proceedings of the SPIE Symposium on Electronic Imaging:Human Vision and Electronic Imaging.SanJose,California,1998,3299-3310
[36]A.H.Gunatilaka and B.A.Baertlein,Feature-level and decision-level fusion of noncoincidently sampled sensors for land mine detection,IEEE Trans.PAMI,23(6)2001.577-589
[37]Baudat G,Anouar F.Kernel-based methods and function approximation[C].Proc of the International Joint Conf on Neural Networks.Piscataway:Institute of Electrical and Electronics Engineers Inc,2001:1244-1249..
[38]廖频.基于统一概率模型的人脸识别技术[博士学位论文].中国科学院研究生院博士论文,2003.
[39]张敏贵,潘泉,张洪才等.多生物特征识别[J].信息与控制,2002,31(6):524-528.
[40]http://www biometriegroup.Eom
[41]陈振洲,李磊,姚正安.基于SVM的特征加权KNN算法[J].中山大学学报(自然科学版),2005,(01)
[42]COVER T M,HART P E.Nearest neighbor pattern classification[J].In Trans IEEE InformTheory,1967,IT-13:21-27.
[43]K.Etemad and T.Chellappa,"Discriminant Analysis for Recognition of Human Face Images,"Journal of the Optical Society of America A,Vol.14,pp.1724-1773,1997.
[44]Klema V C.The Singular Value Decomposition:Its Computation and Some Applications[J].IEEE Transactions on Automatic Control,1980,25(2):164-176.
[45]K.Etemad and T.Chellappa,"Discriminant Analysis for Recognition of Human Face Images,"Journal of the Optical Society of America A,Vol.14,pp.1724-1773,1997.
[46]Li Ming,Yuan Baozong.2D-LDA:A statistical linear discriminant analysis for image matrix.Pattern Recognition Letters,26(2005)5,527-5
[47]Yang Jian,Zhang David,Yong Xu,Yang Jingyu.Two-dimensional discriminant transform for face recognition.Pattern Recognition,38(2005)7,1125-1129.
[48]Kimura F,Shridhar M.Handwritten number recognition based on multiple algorithms.Pattern Recognition,1991,24(10):969-983
[49]A.Nefian,M.H.Hayes Ⅲ.An Embedded HMM-Based Approach for Face Detection and Recognition.Proceedings of IEEE International Conference on Acoustics,Speech,and Signal Processing.1999,6(6):3553-3556.
[50](英)[L.拉宾纳]Lawrence Rabiner,(英)[阮平望]Biing-Hwang Juang著.语音识别基本原理.北京:清华大学出版社,1999.
[51]Samaria F.Face Recognition Using Hidden Markov Models.Engineering Department,Cambridge University.1994.