人脸识别中特征提取与选择算法的研究
|
摘要
人脸识别是计算机视觉研究领域的一个热点问题,并广泛应用于视频监控、门禁系统以及公安系统等领域。本文把混沌理论和人工鱼群算法引入进来,对由LDA变换后的特征空间进行优化搜索,选择出高效低维的人脸鉴别特征,并在此基础上利用MATLAB软件进行仿真实验,验证该方法的正确性和有效性。主要研究工作总结为以下两点：
(1)针对人工鱼群算法易陷入局部最优的问题,提出了一种基于双混沌映射的人工鱼群算法。该方法利用混沌的随机性、遍历性和规律性对人工鱼算法进行改进：一是初始化人工鱼群,增加搜索的多样性；二是在人工鱼陷入局部最优时,对其产生扰动,使其跳出局部最优值,向全局最优值靠近。仿真实验表明：改进后的算法比基本人工鱼群算法的全局寻优能力更强,搜索效率更高。
(2)针对LDA算法无法找到最优分类特征子空间的问题,提出一种基于混沌人工鱼群的LDA算法。该方法把经过LDA算法变换后的特征子空间作为搜索空间,充分利用混沌人工鱼群算法的全局寻优能力,在特征子空间进行优化选择,获得最有利于分类的特征了空间。仿真实验结果表明：混沌人工鱼群与LDA算法的结合,有效地改善了LDA算法在特征向量选择中的局限性,从而提高了识别率。
综上所述,混沌理论、人工鱼群与LDA相结合的方法,对人脸识别中的特征提取与选择具有正确性和高效性,为人脸识别的后续研究奠定了良好的研究基础。
Face recognition technology has been the active field of computer vision, image processing and patern recogniton. After serval decades rapid development, it has been widely applied to surveillance and security, human-computer intelligent interaction, video meeting and so on.This paper persents the method of the feature extraction and selection of face image based on Linear Discriminate Analysis(LDA). The task is the optimization combination of the feature vectors in the feature space after LDA transform by the improved artificial fish school algorithm(AFSA) based on chaos theory. The main task is as follows:
Firstly, a CAFSA algorithm based on chaos theory is presented to resolve the problems of AFSA, such as poor performance of precision and low rate of convergence. Initializing population of fish with the chaos increases the diversity of fish and the chaos search makes fish to get rid of local minima and improve the efficiency. The simulation experiments show that the proposed method has more effective perfomances and robustness.
Secondly, LDA fails to find the optimalest feature subspace for face classification in certain circumstances. Aiming at this problem, an improved LDA algorithm based on CAFSA is proposed in this paper. The novel method LDA-CAFSA selects the best feature from the feature subspace transformed by LDA through the stochastic global optimization ability of the CAFSA. The simulation experiments show that the proposed method is stronger than the basic LDA.
As is stated, the chaos theory and AFSA algorithm with a combination of LDA method is accuracy and credibility algorithm in the extraction and selection of feature space. And this laids a good foundation for the follow study on the face recognition.
(1)针对人工鱼群算法易陷入局部最优的问题,提出了一种基于双混沌映射的人工鱼群算法。该方法利用混沌的随机性、遍历性和规律性对人工鱼算法进行改进：一是初始化人工鱼群,增加搜索的多样性；二是在人工鱼陷入局部最优时,对其产生扰动,使其跳出局部最优值,向全局最优值靠近。仿真实验表明：改进后的算法比基本人工鱼群算法的全局寻优能力更强,搜索效率更高。
(2)针对LDA算法无法找到最优分类特征子空间的问题,提出一种基于混沌人工鱼群的LDA算法。该方法把经过LDA算法变换后的特征子空间作为搜索空间,充分利用混沌人工鱼群算法的全局寻优能力,在特征子空间进行优化选择,获得最有利于分类的特征了空间。仿真实验结果表明：混沌人工鱼群与LDA算法的结合,有效地改善了LDA算法在特征向量选择中的局限性,从而提高了识别率。
综上所述,混沌理论、人工鱼群与LDA相结合的方法,对人脸识别中的特征提取与选择具有正确性和高效性,为人脸识别的后续研究奠定了良好的研究基础。
Face recognition technology has been the active field of computer vision, image processing and patern recogniton. After serval decades rapid development, it has been widely applied to surveillance and security, human-computer intelligent interaction, video meeting and so on.This paper persents the method of the feature extraction and selection of face image based on Linear Discriminate Analysis(LDA). The task is the optimization combination of the feature vectors in the feature space after LDA transform by the improved artificial fish school algorithm(AFSA) based on chaos theory. The main task is as follows:
Firstly, a CAFSA algorithm based on chaos theory is presented to resolve the problems of AFSA, such as poor performance of precision and low rate of convergence. Initializing population of fish with the chaos increases the diversity of fish and the chaos search makes fish to get rid of local minima and improve the efficiency. The simulation experiments show that the proposed method has more effective perfomances and robustness.
Secondly, LDA fails to find the optimalest feature subspace for face classification in certain circumstances. Aiming at this problem, an improved LDA algorithm based on CAFSA is proposed in this paper. The novel method LDA-CAFSA selects the best feature from the feature subspace transformed by LDA through the stochastic global optimization ability of the CAFSA. The simulation experiments show that the proposed method is stronger than the basic LDA.
As is stated, the chaos theory and AFSA algorithm with a combination of LDA method is accuracy and credibility algorithm in the extraction and selection of feature space. And this laids a good foundation for the follow study on the face recognition.
引文
[1]严严,章毓晋.基于视频的人脸识别研究进展[J].计算机学报,2009,32(5)：878-884.
[2]王蕴红,朱勇,谭铁牛.基于虹膜识别的身份鉴定[J].自动化学报,2002,28(1)：1-11.
[3]G.Shakhnarovich, J.W.Fisher, T.Darrell. Face recognition from long-term observations[C]//Proceedings of the 7th European Conference on Computer Vision,2002(3):851-868.
[4]A.K. Jain, A.Ross, S. Prabhakar. An Introduction to Biometric Recognition[J]. Circuitsand Systems for Video Technology,2004, 14(1):4-20.
[5]Zhao W, Chellappa R, Phillips P J, et al. Face recognition:A literature survey[J]. ACM Computing Surveys,2003,35(4):399-459.
[6]Zhou S, Chellappa R. Beyond a single still image:Face recognition from multiple still images and videos[M].New York:Academic Press,2005.
[7]Varma M, Zisserman A. A statistical approach to texture classification from single images [J]. International Journal of ComputerVision,2005,62(12): 61-81.
[8]Arandjelovic'O, Cipolla R. A pose-wise linear illuminationmanifold model for face recognition using video[J]. ComputerVision and Image Understanding,2009,113(1):113-125.
[9]Mikolajczyk K, Schmid C. A performance evaluation of localdescriptors [J]. IEEE Transactions on Pattern Analysis andMachine Intelligence,2005, 27(10):1615-1630.
[10]MujaM, Lowe D G. Fast approximate nearest neighbors with automatic algorithm configuration [C]//Proceedings of the Fourth International Conference on ComputerVision Theory and Applications. New York, USA: ACM Press,2009:331-340.
[11]KeY, Sukthankar. PCA-SIFT:Amore distinctive representationfor local image descriptors [C]//Proceedings of InternationalConference on Computer Vision and Pattern Recognition. NewYork, USA:ACM Press,2004:506-513.
[12]Liu,Y., Mitra, S.Human Identification Versus Expression Classification Via Bagging on Facial Asymmetry[J]. Technical Report CMU-RI-TR-03-08, Carnegie Mellon University,2003,14(8):133-137.
[13]Serre T, Wolf L, Poggio T. Object Recognition with FeaturesInspired by Visual Cortex[C]//Proceedings of IEEE Computer SocietyConference on Computer Vision and Pattern Recognition. SanDiego, USA:IEEE Computer Society,2005: 994-1000.
[14]章毓晋.中国图像工程：2010[J].中国图像图形学报,2011,16(5)：693-702.
[15]胡晓,俞王新,余群等.基于行列特征复融合人脸识别[J].计算机工程,2010,36(11)：176-179.
[16]张凯,苏剑波.基于相似度分布的开集人脸识别方法[J].模式识别与人工智能,2011,24(1)：145-152.
[17]丁嵘,苏光大,林行刚.使用关键点信息改进弹性匹配人脸识别算法[J].电子学报,2000,30(9)：1292-1294.
[18]陈伏兵,杨静宇.分块PCA及其在人脸识别中的应用[J].计算机工程与设计,2007,8(28)：1889-1892.
[19]苏煜,山世光,陈熙霖,高文.基于全局和局部特征集成的人脸识别[J].软件学报,2010,21(8)：1859-1862.
[20]Xu Jian, Ding Xiao-Qing, Wang Sheng-Jin, Wu You-Shou.Background subtraction based on a combination of localtexture and color[J]. Acta Automatica Sinica,2009, 35(9):1145-1150.
[21]吴巾一,周德龙.人脸识别方法综述[J].计算机应用研究,2009,26(9)：3205-3209.
[22]陈伏兵.人脸识别中鉴别特征抽取若干方法研究[D].中国优秀博士论文,2006.
[23]杨万扣,王建国,任明武等.模糊逆Fisher鉴别分析及其在人脸识别中的应用[J].中国图象图形学报,2009,14(1)：88-93.
[24]刘青山,卢汉清,马颂德.综述人脸识别中子空间方法[J].自动化学报,2003,29(6)：900-911.
[25]王展青,刘小双,张桂林等.基于PCA与ICA的人脸识别算法研究[J].华中师范大学学报(自然科学版),2007,41(3)：373-376.
[26]郝红卫,张蕾.基于SVD和LDA的人脸识别方法[J].计算机应用研究,2007,24(12)：377-392.
[27]谢永林.LDA算法及其在人脸识别中的应用[J].计算机工程与应用,2010,46(19)：189-192
[28]王建国,杨万扣,郑宇杰等.一种基于ICA和模糊LDA的特征提取方法[J].模式识别与人工智能,2008,21(6)：819-823
[29]朱玉莲.半随机子空间的LDA人脸识别方法[J].计算机工程与应用,2010, 46(20)：197-201.
[30]L. Chen, H. Liao, M. Ko, et al. A new LDA-based face recognition which can solve the small sample size problem [J]. Pattern Recognition,2000,33(10):1713-1726.
[31]H. Yu, J. Yang. A direct LDA algorithm for high-dimensional data-with application to face recognition [J]. Pattern Recognition,2001,34(10):2067-2070.
[32]刘永俊,陈才扣,王正群.修正的最大散度差鉴别分析及人脸识别[J].电子与信息学报,2008,30(1)：190-195.
[33]陈才扣,刘永俊,杨静宇.二维最大散度差图像投影鉴别分析[J].系统仿真学报,2007,19(14)：833-837.
[34]ZHENG W M, ZOU C R,ZHAO L. An improved algorithm for kernel principal component analysis[J].Neural Processing Letters,2005,22(1):49-56.
[35]周德龙.人脸识别技术研究[D].西安：西北工业大学,2001.
[36]J. Yang, D. Zhang, J. Y. Yang, B. Niu. Globally Maximizing, Locally Minimizing: Unsupervised Discriminant Projection with Applications to Face and Palm Biometrics [J]. Pattern Anal and Mach Intell,2007,29(4):650-664.
[37]X. He, S. Yan, Y. Hu, Niyogi, H. Zhang. Face Recognition Using Laplacianfaces [J]. Pattern Analysis and Machine Intelligence.2005,27(3):328-340.
[38]S. Yah, D. Xu, B. Zhang, and H. J. Zhang. Graph Embedding and Extensions:A General Framework for Dimensionality Reduction [J]. Pattern Anal and Mach Intell, 2007,29(1):40-51.
[39]H. T. Chen, H. W. Chang, and T. L. Liu. Local Disciminant Embedding and Its Variants[C]//Proceedings of Computer Vision and Pattern Recognition. 2005:846-853.
[40]张祥德,张大为,唐青松等.仿生算法与主成分分析相融合的人脸识别方法[J].东北大学学报(自然科学版),2009,30(7)：972-974.
[41]曲良东,何登旭.一种混沌人工鱼群优化算法[J].计算机工程与应用.2010,46(22)：40-42.
[42]李晓磊,路飞,田国会,等.组合优化问题的人工鱼群算法应用[J].山东大学学报,2004,34(5)：65-68.
[43]李晓磊,钱积新.基于分解协调的人工鱼群优化算法研究[J].电路与系统学报,2003,8(1)：1-6.
[44]李晓磊,邵之江,钱积新.一种基于动物自治体的寻优模式：鱼群算法[J].系统工程理论与实践,2002,22(11)：32-38.
[45]周茜.混沌理论及应用若干问题的研究[D].上海：南开大学,2010.
[46]Zeraoulia E, Sprott J C. On the robustness of chaos in dynamical systems:Theories and applications[J]. Frontiers of Physics in China,2008, 3(2):195-204.
[47]May R. Simple mathematical models with very complicated dynamics[J]. Nature, 261,1976.
[48]王凌,郑大钟,李清生.混沌优化算法的研究进展[J].计算技术与自动化,2001,20(1)：1-5.
[49]温浩,郭崇慧.利用粒子群优化的人脸特征提取识别算法[J].西安交通大学学报,2010,44(4)：48-51.
[50]徐春明.一种基于PSO权重优化的加权二维主分量分析方法[J].大连民族学院学报,2009,11(3)：264-268.
[51]WANG Ling,YU Jinshou. Fault Feature selection based onmodified binaryPSO withmutation and its application in chemical process faultdiagnosis[J]. Lecture notes in computer science,2005,3612(7): 832-840.
[52]张宗飞.一种改进型量子遗传算法[J].计算机工程,2010,36(6)：181-185.
[53]黄美灵,赵之杰,浦立娜等.基于自适应Tent混沌搜索的粒子群优化算法[J].计算机应用,2011,31(2)：485-491.
[54]王建国,杨万扣,郑宇杰,杨静宇.一种基于ICA和模糊LDA的特征提取方法[J].模式识别与人工智能,2008,21(6)：819-823.
[55]L. Chen, H. Liao, M. Ko, et al. A new LDA-based face recognition which can solve the small sample size problem [J]. Pattern Recognition,2000,33(10):1713-1726.
[56]Peter N.Belhumeur, et al. Eigenfaces vs.Fisherfaces:Recognition using class specific linear projection[J]. Pattern Anal.Machine Intell.1997, 19(7):711-720
[57]Chen Li-Fen, Liao H-Y Mark, Ko M-T, et al.A new LDA-based face recognition system which can solve the small sample size problem[J]. Pattern Recogniton 2000,33(10):1713-1726.
[58]Yu Hua, YANG Jie. A direct LDA algorithm for high-dimensional data-with application to face recogniton[J]. Pattern Recogniton,2001,34(10):2067-2070.
[59]赵武锋,沈海斌,严晓浪.直接LDA在人脸识别中的鉴别力分析[J].浙江大学学报2010,44(8)：1479-1483.
[60]Paul M B.The class-specific classifier:avoiding the curse of dimensionality[J]. Aerospace and Electronic Systems Magazine,2004,19(1):37-52.
[61]Caputo B, Niemann H. To each according to its need:kernel class specific classifiers[C]//Proceedings of the 16th International Conference on Pattern Recognition. Quebec. Canada:IEEE Press,2002:94-97.
[62]陈晓红.类依赖的相关性多类分类器[J].计算机工程与应用,2010,46(2)：7-10.
[63]赵武锋.人脸识别中特征提取方法的研究[D].‘浙江：浙江大学,2009.
[2]王蕴红,朱勇,谭铁牛.基于虹膜识别的身份鉴定[J].自动化学报,2002,28(1)：1-11.
[3]G.Shakhnarovich, J.W.Fisher, T.Darrell. Face recognition from long-term observations[C]//Proceedings of the 7th European Conference on Computer Vision,2002(3):851-868.
[4]A.K. Jain, A.Ross, S. Prabhakar. An Introduction to Biometric Recognition[J]. Circuitsand Systems for Video Technology,2004, 14(1):4-20.
[5]Zhao W, Chellappa R, Phillips P J, et al. Face recognition:A literature survey[J]. ACM Computing Surveys,2003,35(4):399-459.
[6]Zhou S, Chellappa R. Beyond a single still image:Face recognition from multiple still images and videos[M].New York:Academic Press,2005.
[7]Varma M, Zisserman A. A statistical approach to texture classification from single images [J]. International Journal of ComputerVision,2005,62(12): 61-81.
[8]Arandjelovic'O, Cipolla R. A pose-wise linear illuminationmanifold model for face recognition using video[J]. ComputerVision and Image Understanding,2009,113(1):113-125.
[9]Mikolajczyk K, Schmid C. A performance evaluation of localdescriptors [J]. IEEE Transactions on Pattern Analysis andMachine Intelligence,2005, 27(10):1615-1630.
[10]MujaM, Lowe D G. Fast approximate nearest neighbors with automatic algorithm configuration [C]//Proceedings of the Fourth International Conference on ComputerVision Theory and Applications. New York, USA: ACM Press,2009:331-340.
[11]KeY, Sukthankar. PCA-SIFT:Amore distinctive representationfor local image descriptors [C]//Proceedings of InternationalConference on Computer Vision and Pattern Recognition. NewYork, USA:ACM Press,2004:506-513.
[12]Liu,Y., Mitra, S.Human Identification Versus Expression Classification Via Bagging on Facial Asymmetry[J]. Technical Report CMU-RI-TR-03-08, Carnegie Mellon University,2003,14(8):133-137.
[13]Serre T, Wolf L, Poggio T. Object Recognition with FeaturesInspired by Visual Cortex[C]//Proceedings of IEEE Computer SocietyConference on Computer Vision and Pattern Recognition. SanDiego, USA:IEEE Computer Society,2005: 994-1000.
[14]章毓晋.中国图像工程：2010[J].中国图像图形学报,2011,16(5)：693-702.
[15]胡晓,俞王新,余群等.基于行列特征复融合人脸识别[J].计算机工程,2010,36(11)：176-179.
[16]张凯,苏剑波.基于相似度分布的开集人脸识别方法[J].模式识别与人工智能,2011,24(1)：145-152.
[17]丁嵘,苏光大,林行刚.使用关键点信息改进弹性匹配人脸识别算法[J].电子学报,2000,30(9)：1292-1294.
[18]陈伏兵,杨静宇.分块PCA及其在人脸识别中的应用[J].计算机工程与设计,2007,8(28)：1889-1892.
[19]苏煜,山世光,陈熙霖,高文.基于全局和局部特征集成的人脸识别[J].软件学报,2010,21(8)：1859-1862.
[20]Xu Jian, Ding Xiao-Qing, Wang Sheng-Jin, Wu You-Shou.Background subtraction based on a combination of localtexture and color[J]. Acta Automatica Sinica,2009, 35(9):1145-1150.
[21]吴巾一,周德龙.人脸识别方法综述[J].计算机应用研究,2009,26(9)：3205-3209.
[22]陈伏兵.人脸识别中鉴别特征抽取若干方法研究[D].中国优秀博士论文,2006.
[23]杨万扣,王建国,任明武等.模糊逆Fisher鉴别分析及其在人脸识别中的应用[J].中国图象图形学报,2009,14(1)：88-93.
[24]刘青山,卢汉清,马颂德.综述人脸识别中子空间方法[J].自动化学报,2003,29(6)：900-911.
[25]王展青,刘小双,张桂林等.基于PCA与ICA的人脸识别算法研究[J].华中师范大学学报(自然科学版),2007,41(3)：373-376.
[26]郝红卫,张蕾.基于SVD和LDA的人脸识别方法[J].计算机应用研究,2007,24(12)：377-392.
[27]谢永林.LDA算法及其在人脸识别中的应用[J].计算机工程与应用,2010,46(19)：189-192
[28]王建国,杨万扣,郑宇杰等.一种基于ICA和模糊LDA的特征提取方法[J].模式识别与人工智能,2008,21(6)：819-823
[29]朱玉莲.半随机子空间的LDA人脸识别方法[J].计算机工程与应用,2010, 46(20)：197-201.
[30]L. Chen, H. Liao, M. Ko, et al. A new LDA-based face recognition which can solve the small sample size problem [J]. Pattern Recognition,2000,33(10):1713-1726.
[31]H. Yu, J. Yang. A direct LDA algorithm for high-dimensional data-with application to face recognition [J]. Pattern Recognition,2001,34(10):2067-2070.
[32]刘永俊,陈才扣,王正群.修正的最大散度差鉴别分析及人脸识别[J].电子与信息学报,2008,30(1)：190-195.
[33]陈才扣,刘永俊,杨静宇.二维最大散度差图像投影鉴别分析[J].系统仿真学报,2007,19(14)：833-837.
[34]ZHENG W M, ZOU C R,ZHAO L. An improved algorithm for kernel principal component analysis[J].Neural Processing Letters,2005,22(1):49-56.
[35]周德龙.人脸识别技术研究[D].西安：西北工业大学,2001.
[36]J. Yang, D. Zhang, J. Y. Yang, B. Niu. Globally Maximizing, Locally Minimizing: Unsupervised Discriminant Projection with Applications to Face and Palm Biometrics [J]. Pattern Anal and Mach Intell,2007,29(4):650-664.
[37]X. He, S. Yan, Y. Hu, Niyogi, H. Zhang. Face Recognition Using Laplacianfaces [J]. Pattern Analysis and Machine Intelligence.2005,27(3):328-340.
[38]S. Yah, D. Xu, B. Zhang, and H. J. Zhang. Graph Embedding and Extensions:A General Framework for Dimensionality Reduction [J]. Pattern Anal and Mach Intell, 2007,29(1):40-51.
[39]H. T. Chen, H. W. Chang, and T. L. Liu. Local Disciminant Embedding and Its Variants[C]//Proceedings of Computer Vision and Pattern Recognition. 2005:846-853.
[40]张祥德,张大为,唐青松等.仿生算法与主成分分析相融合的人脸识别方法[J].东北大学学报(自然科学版),2009,30(7)：972-974.
[41]曲良东,何登旭.一种混沌人工鱼群优化算法[J].计算机工程与应用.2010,46(22)：40-42.
[42]李晓磊,路飞,田国会,等.组合优化问题的人工鱼群算法应用[J].山东大学学报,2004,34(5)：65-68.
[43]李晓磊,钱积新.基于分解协调的人工鱼群优化算法研究[J].电路与系统学报,2003,8(1)：1-6.
[44]李晓磊,邵之江,钱积新.一种基于动物自治体的寻优模式：鱼群算法[J].系统工程理论与实践,2002,22(11)：32-38.
[45]周茜.混沌理论及应用若干问题的研究[D].上海：南开大学,2010.
[46]Zeraoulia E, Sprott J C. On the robustness of chaos in dynamical systems:Theories and applications[J]. Frontiers of Physics in China,2008, 3(2):195-204.
[47]May R. Simple mathematical models with very complicated dynamics[J]. Nature, 261,1976.
[48]王凌,郑大钟,李清生.混沌优化算法的研究进展[J].计算技术与自动化,2001,20(1)：1-5.
[49]温浩,郭崇慧.利用粒子群优化的人脸特征提取识别算法[J].西安交通大学学报,2010,44(4)：48-51.
[50]徐春明.一种基于PSO权重优化的加权二维主分量分析方法[J].大连民族学院学报,2009,11(3)：264-268.
[51]WANG Ling,YU Jinshou. Fault Feature selection based onmodified binaryPSO withmutation and its application in chemical process faultdiagnosis[J]. Lecture notes in computer science,2005,3612(7): 832-840.
[52]张宗飞.一种改进型量子遗传算法[J].计算机工程,2010,36(6)：181-185.
[53]黄美灵,赵之杰,浦立娜等.基于自适应Tent混沌搜索的粒子群优化算法[J].计算机应用,2011,31(2)：485-491.
[54]王建国,杨万扣,郑宇杰,杨静宇.一种基于ICA和模糊LDA的特征提取方法[J].模式识别与人工智能,2008,21(6)：819-823.
[55]L. Chen, H. Liao, M. Ko, et al. A new LDA-based face recognition which can solve the small sample size problem [J]. Pattern Recognition,2000,33(10):1713-1726.
[56]Peter N.Belhumeur, et al. Eigenfaces vs.Fisherfaces:Recognition using class specific linear projection[J]. Pattern Anal.Machine Intell.1997, 19(7):711-720
[57]Chen Li-Fen, Liao H-Y Mark, Ko M-T, et al.A new LDA-based face recognition system which can solve the small sample size problem[J]. Pattern Recogniton 2000,33(10):1713-1726.
[58]Yu Hua, YANG Jie. A direct LDA algorithm for high-dimensional data-with application to face recogniton[J]. Pattern Recogniton,2001,34(10):2067-2070.
[59]赵武锋,沈海斌,严晓浪.直接LDA在人脸识别中的鉴别力分析[J].浙江大学学报2010,44(8)：1479-1483.
[60]Paul M B.The class-specific classifier:avoiding the curse of dimensionality[J]. Aerospace and Electronic Systems Magazine,2004,19(1):37-52.
[61]Caputo B, Niemann H. To each according to its need:kernel class specific classifiers[C]//Proceedings of the 16th International Conference on Pattern Recognition. Quebec. Canada:IEEE Press,2002:94-97.
[62]陈晓红.类依赖的相关性多类分类器[J].计算机工程与应用,2010,46(2)：7-10.
[63]赵武锋.人脸识别中特征提取方法的研究[D].‘浙江：浙江大学,2009.