基于Adaboost的BCI系统脑电信号分类
|
摘要
脑机接口是建立在人脑和外部设备之间的直接通讯通路。脑机接口技术结合了生物医学、计算机信息处理技术、神经科学以及微电子等多个领域的最新成果,在近10年的时间里得到的广泛的重视研究和发展。通过对脑电信号的研究,脑机接口BCI(Brain-Computer Systerm)系统已经可以被用来解决很多实际问题。尽管如此,因为脑电信号的不稳定性和个体差异性,找到一种高效的、具有普遍意义的信号处理和识别方式是解决问题的关键。本文的主要研究对象是BCI系统中脑电信号的处理和识别方法。通常一个模式识别过程可以分为数据预处理、特征提取、特征选择和降维、以及特征分类几个阶段。在脑电信号的预处理阶段使用FIR数字滤波器和CSP空域滤波器的方法对脑电信号进行滤波处理,用主成分分析PCA和偏最小二乘PLS的方法对脑电信号的特征向量进行降维,用两种不同的Adaboost分类方法对提取到的脑电信号特征向量进行分类,同时用线性判别分写LDA和支持向量机SVM的BCI系统经典分类方法进行分类识别率的对比。
通过对在模式识别每一个阶段对脑电信号处理的不同方法的研究,在实验结果数据的基础上给出一种高效的处理脑电信号的方法,即公共空间模式滤波与以最近邻法为若分类其的AdaboostNN分类器结合的处理方式。
Brain-machine interface is direct communication access which is based on the human brain and the external device. Brain-machine interface technology combines the latest achievements of bio medicine, computer information processing technology, neuroscience, microelectronics and other fields, and in the past 10 years, it has developed quickly. With the study of EEG,BCI system can already be used to solve many practical problems. Because of the instability of EEG and individual differences, people need to find an efficient and stable signal processing and recognition approach in BCI systerm. This paper studies the EEG processing and recognition methods of the BCI system. Usually a pattern recognition process can be divided into data preprocessing, feature extraction, feature selection and dimensionality reduction, and feature classification. In the stage of EEG preprocessing, using FIR digital filters and CSP spatial filter for EEG filter processing. Using principal component analysis PCA and partial least squares PLS to reduce dimensions of EEG feature vectors. Classification methods used here are two different Adaboost classification methods for classifying the extracted feature vectors of EEG signals. Linear discriminant analysis LDA and support vector machine SVM which are classic classification methods of BCI system are applied for comparison of the recognition rate.
EEG signals are processed in a BCI system by those different methods above, based on the experimental results,this paper gives a efficient EEG signal processing method,i.e. a combination method of CSP filtering and AdaboostNN classifier.
通过对在模式识别每一个阶段对脑电信号处理的不同方法的研究,在实验结果数据的基础上给出一种高效的处理脑电信号的方法,即公共空间模式滤波与以最近邻法为若分类其的AdaboostNN分类器结合的处理方式。
Brain-machine interface is direct communication access which is based on the human brain and the external device. Brain-machine interface technology combines the latest achievements of bio medicine, computer information processing technology, neuroscience, microelectronics and other fields, and in the past 10 years, it has developed quickly. With the study of EEG,BCI system can already be used to solve many practical problems. Because of the instability of EEG and individual differences, people need to find an efficient and stable signal processing and recognition approach in BCI systerm. This paper studies the EEG processing and recognition methods of the BCI system. Usually a pattern recognition process can be divided into data preprocessing, feature extraction, feature selection and dimensionality reduction, and feature classification. In the stage of EEG preprocessing, using FIR digital filters and CSP spatial filter for EEG filter processing. Using principal component analysis PCA and partial least squares PLS to reduce dimensions of EEG feature vectors. Classification methods used here are two different Adaboost classification methods for classifying the extracted feature vectors of EEG signals. Linear discriminant analysis LDA and support vector machine SVM which are classic classification methods of BCI system are applied for comparison of the recognition rate.
EEG signals are processed in a BCI system by those different methods above, based on the experimental results,this paper gives a efficient EEG signal processing method,i.e. a combination method of CSP filtering and AdaboostNN classifier.
引文
[1]王磊.基于运动想象的脑电信号分类与脑机接口技术研究.博士,河北工业大学,2008.
[2]高上凯.浅谈脑—机接口的发展现状与挑战.中国生物医学工程学报,2007,26(6)：801-803
[3]G Pfurtscheller, C. Neuper. Motor imagery activates primary sensorimotor area in humans, Neuroscience Letters.1997,239:65-68.
[4]P. Ritter, M. Moosmann. A. Villringer, Rolandic alpha and beta EEG rhythms' strengths are inversely related to fMRI-BOLD signal in primary somatosensory and motor cortex. Human Brain Mapping,2007,30:1168-1187.
[5]Grose-Fifer Jillian, Deacon Diana. Priming by natural category membership in the left and right cerebral hemispheres, Neuropsychologia.2004,42(14):1948-1960.
[6]Wolpaw JR, Birbaumer N, McFarland WJ et al. Brain-computer interface technology:A review of the first international meeting, IEEE Transactions on Neural Systems and Rehabilitation Engineering.2000,8(2)164-173.
[7]P.S. Hammon, V.R de Sa. Preprocessing and meta-classification for Brain-Computer Interfaces, IEEE Transactions on Biomedical Engineering.2007, 54:518-525.
[8]B. Blankertz, R. Tomioka, S. Lemm, et al. Optimizing spatial filters for robust EEG single-trial analysis, IEEE Signal Processing Magazine.2008,25:41-56.
[9]J. MUller-Gerking, G Pfurtscheller, and H. Flyvbjerg. Designing optimal spatial filters for single-trial EEG classification in a movement task. Clinical Neurophysiology,1999,110:787-798.
[10]Sinkjaer Thomas, Haugland Morten, Inmann Andreas. Biopotentials as command and feedback signals in functional electrical stimulation systems. Medical Engineering and Physics,2003,25(1):29-40.
[11]Konstuntions I P, Leontios I H, Stavros M P. Hilbert-Huang Spectrum as a New field for Identification of EEG Event Related De/synchronization for BCI application. Proc. of the 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Vancouver, Canada,2008, 3832-3835.
[12]V.Vapnik.统计学习理论的本质.北京：清华大学出版社,2000.
[13]王惠文,吴载斌,孟洁.偏最小二乘回归方法及其应用,北京：国防工业出版,1999.
[14]边肇祺,张学工.模式识别(第二版).北京：清华大学出版社,2004.
[15]王兴玲,李占斌.基于网格搜索的支持向量机核函数参数的确定,中国海洋 大学学报,2005,35(5)：859-862
[16]M. Asghari Oskoei, J.Q. Gan, and H. Hu. Adaptive schemes applied to online SVM for BCI data classification,31st Annual International Conference of the IEEE EMBS Minneapolis,2009.
[17]Maryam Mohebbi, Hassan Ghassemian, Detection of Atrial Fibrillation Episodes Using SVM, Proc. of the 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society,2008:177-180.
[18]Lei Wang, Gui Zhi Xu, Lei Guo, et al.3D Reconstruction of Head MRI Based on One Class Support Vector Machine with Immune Algorithm, Proc. of the 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society,2007:6015-6018.
[19]Yoav Freund, Robert E.Schapire. A decision-theoretic generalization of on-line learning and an application to boosting. Computational Learning Theory, 1995:23-37.
[20]R. E. Schapire, Y Freund, P. Bartlett et al. Boosting the margin:a new explanation for the effectiveness of voting methods. Ann. Stat,1998, 26(5):1651-1686.
[21]P. Viola, M. Jones, Robust Real-time Object Detection, IEEE ICCV Workshop on Statistical and Computational Theories of Vision,2001.
[22]Chunhua Shen and Hanxi Li, On the dual formulation of Boosting algorithms, IEEE Transactions on Pattern Analysis and Machine Intelligence,2010, 32:2216-2231.
[23]R.E Schapire and Y Singer. Improved boosting algorithms using confidence-rated predictions. Machine Learning,1999,37:297-336.
[24]J. Friedman, T. Hastie, and R. Tibshirani. Additive logistic regression:a statistical view of boosting. Annals of Statistics,2000,28:337-374.
[25]Devroye, L., Gyorfi, L., Krzyzak, A., et al. On the strong universal consistency of nearest neighbor regression function estimates. Annals of Statistics, 22:1371-1385.
[26]B. Blankertz. BCI Competition Ⅲ, Fraunhofer FIRST (IDA), http://ida.first. fraunhofer.de/projects/bci/competition_ⅲ/,2005.
[27]B. Blankertz, BCI Competition IV, Fraunhofer FIRST (IDA), http://ida.first.fraunhofer.de/projects/bci/competition_ⅳ/,2008.
[2]高上凯.浅谈脑—机接口的发展现状与挑战.中国生物医学工程学报,2007,26(6)：801-803
[3]G Pfurtscheller, C. Neuper. Motor imagery activates primary sensorimotor area in humans, Neuroscience Letters.1997,239:65-68.
[4]P. Ritter, M. Moosmann. A. Villringer, Rolandic alpha and beta EEG rhythms' strengths are inversely related to fMRI-BOLD signal in primary somatosensory and motor cortex. Human Brain Mapping,2007,30:1168-1187.
[5]Grose-Fifer Jillian, Deacon Diana. Priming by natural category membership in the left and right cerebral hemispheres, Neuropsychologia.2004,42(14):1948-1960.
[6]Wolpaw JR, Birbaumer N, McFarland WJ et al. Brain-computer interface technology:A review of the first international meeting, IEEE Transactions on Neural Systems and Rehabilitation Engineering.2000,8(2)164-173.
[7]P.S. Hammon, V.R de Sa. Preprocessing and meta-classification for Brain-Computer Interfaces, IEEE Transactions on Biomedical Engineering.2007, 54:518-525.
[8]B. Blankertz, R. Tomioka, S. Lemm, et al. Optimizing spatial filters for robust EEG single-trial analysis, IEEE Signal Processing Magazine.2008,25:41-56.
[9]J. MUller-Gerking, G Pfurtscheller, and H. Flyvbjerg. Designing optimal spatial filters for single-trial EEG classification in a movement task. Clinical Neurophysiology,1999,110:787-798.
[10]Sinkjaer Thomas, Haugland Morten, Inmann Andreas. Biopotentials as command and feedback signals in functional electrical stimulation systems. Medical Engineering and Physics,2003,25(1):29-40.
[11]Konstuntions I P, Leontios I H, Stavros M P. Hilbert-Huang Spectrum as a New field for Identification of EEG Event Related De/synchronization for BCI application. Proc. of the 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Vancouver, Canada,2008, 3832-3835.
[12]V.Vapnik.统计学习理论的本质.北京：清华大学出版社,2000.
[13]王惠文,吴载斌,孟洁.偏最小二乘回归方法及其应用,北京：国防工业出版,1999.
[14]边肇祺,张学工.模式识别(第二版).北京：清华大学出版社,2004.
[15]王兴玲,李占斌.基于网格搜索的支持向量机核函数参数的确定,中国海洋 大学学报,2005,35(5)：859-862
[16]M. Asghari Oskoei, J.Q. Gan, and H. Hu. Adaptive schemes applied to online SVM for BCI data classification,31st Annual International Conference of the IEEE EMBS Minneapolis,2009.
[17]Maryam Mohebbi, Hassan Ghassemian, Detection of Atrial Fibrillation Episodes Using SVM, Proc. of the 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society,2008:177-180.
[18]Lei Wang, Gui Zhi Xu, Lei Guo, et al.3D Reconstruction of Head MRI Based on One Class Support Vector Machine with Immune Algorithm, Proc. of the 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society,2007:6015-6018.
[19]Yoav Freund, Robert E.Schapire. A decision-theoretic generalization of on-line learning and an application to boosting. Computational Learning Theory, 1995:23-37.
[20]R. E. Schapire, Y Freund, P. Bartlett et al. Boosting the margin:a new explanation for the effectiveness of voting methods. Ann. Stat,1998, 26(5):1651-1686.
[21]P. Viola, M. Jones, Robust Real-time Object Detection, IEEE ICCV Workshop on Statistical and Computational Theories of Vision,2001.
[22]Chunhua Shen and Hanxi Li, On the dual formulation of Boosting algorithms, IEEE Transactions on Pattern Analysis and Machine Intelligence,2010, 32:2216-2231.
[23]R.E Schapire and Y Singer. Improved boosting algorithms using confidence-rated predictions. Machine Learning,1999,37:297-336.
[24]J. Friedman, T. Hastie, and R. Tibshirani. Additive logistic regression:a statistical view of boosting. Annals of Statistics,2000,28:337-374.
[25]Devroye, L., Gyorfi, L., Krzyzak, A., et al. On the strong universal consistency of nearest neighbor regression function estimates. Annals of Statistics, 22:1371-1385.
[26]B. Blankertz. BCI Competition Ⅲ, Fraunhofer FIRST (IDA), http://ida.first. fraunhofer.de/projects/bci/competition_ⅲ/,2005.
[27]B. Blankertz, BCI Competition IV, Fraunhofer FIRST (IDA), http://ida.first.fraunhofer.de/projects/bci/competition_ⅳ/,2008.