脑电特征提取与在线脑机接口研究
|
摘要
脑机接口BCI是一种新型的人机交互方式，它不依赖于大脑的正常输出通道（外围神经和肌肉组织等），实现人脑直接与外部环境的信息交互。BCI作为一种交叉学科技术，涉及到医学工程、康复工程、电子工程及人工智能等领域。BCI系统解读大脑思维，正确认识大脑各源与思维动作之间的关系尤其重要，其中脑电信号EEG的特征提取、模式分类理论方法的研究一直是热点问题。
本文针对EEG信号的特征提取方法做了离线分析研究，在已有研究的基础上作了如下工作：
首先，针对传统的线性判别分析中小样本采样问题，以及其核推广核线性判别分析，提出了一种基于广义奇异值分解的广义核线性判别分析。为了最大限度的释放不同类之间的非线性模式，同时解决小样本采样问题，采用一种广义奇异值分解的方法，解除了奇异性限定。
其次，同样针对小样本采样问题，本文将鉴别公共矢量方法应用在脑机接口特征提取中。在脑机接口中采集的脑电信号，其样本数是远远小于或等于样本空间的维数的，也就是通常所谓的小样本采样问题。鉴别公共矢量方法中，借鉴了公共矢量的理论，即通过消除类内离散矩阵的非零特征值所对应的特征向量方向上的共有特征，该公共矢量可作为分类的特征。
最后，本文针对在线BCI做了研究，实现了基于P300电位的脑机接口小车控制。本文对P300电位做了深入研究，并在BCI2000平台下，加入了外设控制命令，成功的完成在线视觉刺激BCI外部设备控制系统。受试者几乎不需要经过训练，或小量训练，便可通过视觉刺激来控制小车的动作。
Brain-Computer Interface is a new way of human-computer interface.It is notdependent on the brain’s normal output channels (peripheral nerves and muscles),butdirectly change the information between the brain and outside world.As amultidisciplinary cross technology, BCI relate to Biomedical Engineering, RehabilitationEngineering, Electronics Engineering and Artificial Intelligence.BCI system can explainthe brain’ mental task, it is vary important to get the relationship between brain’s sourceand mental task, The theory research such as feature extraction, classification andexperiment research of electroencephalogram(EEG) play important roles.
EEG signal can be collected conveniently, so it can be easily popularize. The maincontributions of the paper are as follows:
First of all, a new method which combines GSVD and kernel method is proposed tosolve the problems of small sample size in traditional LDA and kernel LDA. In order torelax the presumption of strictly linear pattern and relieve the nonsingular, GSVD is usedto solve the optimizing spatial filter.
Secondly, for the small sample size, this paper use another method calledDiscriminative Common Vector. In BCI system, the number of sample of EEG signal islow than the sample space dimension, i.e. small sample size problem. DCV removes allthe features that are in the direction of the eigenvectors corresponding to the nonzeroeigenvalues of the scatter matrix. The DCV can be the feature to be classified.
At last, the online BCI is involve. A BCI car controlled system based on P300potential is realized. We makes a deep research on P300potential. With the BCI2000platform, an online visual stimulation BCI external equipment control system is realized.The research subjects can control the car by visual stimulation after a short train time.
本文针对EEG信号的特征提取方法做了离线分析研究，在已有研究的基础上作了如下工作：
首先，针对传统的线性判别分析中小样本采样问题，以及其核推广核线性判别分析，提出了一种基于广义奇异值分解的广义核线性判别分析。为了最大限度的释放不同类之间的非线性模式，同时解决小样本采样问题，采用一种广义奇异值分解的方法，解除了奇异性限定。
其次，同样针对小样本采样问题，本文将鉴别公共矢量方法应用在脑机接口特征提取中。在脑机接口中采集的脑电信号，其样本数是远远小于或等于样本空间的维数的，也就是通常所谓的小样本采样问题。鉴别公共矢量方法中，借鉴了公共矢量的理论，即通过消除类内离散矩阵的非零特征值所对应的特征向量方向上的共有特征，该公共矢量可作为分类的特征。
最后，本文针对在线BCI做了研究，实现了基于P300电位的脑机接口小车控制。本文对P300电位做了深入研究，并在BCI2000平台下，加入了外设控制命令，成功的完成在线视觉刺激BCI外部设备控制系统。受试者几乎不需要经过训练，或小量训练，便可通过视觉刺激来控制小车的动作。
Brain-Computer Interface is a new way of human-computer interface.It is notdependent on the brain’s normal output channels (peripheral nerves and muscles),butdirectly change the information between the brain and outside world.As amultidisciplinary cross technology, BCI relate to Biomedical Engineering, RehabilitationEngineering, Electronics Engineering and Artificial Intelligence.BCI system can explainthe brain’ mental task, it is vary important to get the relationship between brain’s sourceand mental task, The theory research such as feature extraction, classification andexperiment research of electroencephalogram(EEG) play important roles.
EEG signal can be collected conveniently, so it can be easily popularize. The maincontributions of the paper are as follows:
First of all, a new method which combines GSVD and kernel method is proposed tosolve the problems of small sample size in traditional LDA and kernel LDA. In order torelax the presumption of strictly linear pattern and relieve the nonsingular, GSVD is usedto solve the optimizing spatial filter.
Secondly, for the small sample size, this paper use another method calledDiscriminative Common Vector. In BCI system, the number of sample of EEG signal islow than the sample space dimension, i.e. small sample size problem. DCV removes allthe features that are in the direction of the eigenvectors corresponding to the nonzeroeigenvalues of the scatter matrix. The DCV can be the feature to be classified.
At last, the online BCI is involve. A BCI car controlled system based on P300potential is realized. We makes a deep research on P300potential. With the BCI2000platform, an online visual stimulation BCI external equipment control system is realized.The research subjects can control the car by visual stimulation after a short train time.
引文
[1] Wolpaw J R, Birbaumer N, McFarland D J, et al. Brain-Computer Interfaces for Communicationand Control[J]. Clinical Neurophysiology,2002,113(6):767–791.
[2]王毅军.基于节律调制的脑－机接口系统—从离线到在线的跨越[D].北京：清华大学生物医学工程学科博士学位论文，2007:14–16.
[3] Schalk G, McFarland D J, Hinterberger T,et al. BCI2000:A General-Purpose Brain-ComputerInterface(BCI)System[C]. IEEE Trans Biomed Eng.2004,51(06):1034–1043.
[4] Schalk G, Jürgen M. A Pratical Guide to Brain-Computer Interfacing with BCI2000[M]. SpringerLondon Dordrecht Heidelberg New York.2010:4–5.
[5]王金甲,周丽娜,赵玉超.基于MEG的脑机接口特征提取方法研究[J].仪器仪表学报,2010,31(7):1460–1465.
[6] Birch G E, Mason S G, Borisoff J F. Current Trends in Brain-Computer Interface Research at theNeil Squire Foundation[J]. IEEE Transaction on Neural Systems and Rehabilitation Engineering,2003,11(2):123–126.
[7] Pfurtscheller G, Neuper C. Graze-BCI:State of the Art and Clinical Applications[J]. IEEETransaction on Neural Systems and Rehabilitation Engineering,2003,11(2):177–180.
[8] Gao X R, Xu D F, Cheng M. A BCI-based Environmental Controller for the Motiondisabled[J].IEEE Transaction on Neural Systems and Rehabilitation Engineering,2003,11(2):137–140.
[9]边肇祺,张学工.模式识别[M].北京:清华大学出版社,2000:84–90.
[10]马忠伟,高上凯.基于P300的脑-机接口：视觉刺激强度对性能的影响[J].清华大学学报,2008,48(3):415–418.
[11] Hoffmann U, Vesin J M, Ebrahimi T, et al. An Efficient P300-based Brain-Computer Interface forDisabled Subjects[J]. Journal of Neuroscience Methods.2008,167(1):115–125.
[12] Finke A, Lenhardt A, Ritter H. The MindGame: A P300-based Brain-Computer Interface Game[J].Neural Networks.2009,22(9):1329–1333.
[13] Zhang H, Guan C, Wang C. Asynchronous P300-Based Brain--Computer Interfaces: AComputational Approach With Statistical Models[J]. Biomedical Engineering, IEEE Transactionson,2008,55(6):1754–176.
[14]吴边,苏煜,张剑慧,等.基于P300电位的新型BCI中文输入虚拟键盘系统[J].电子学报,2009,37(8):1733–1738.
[15] Furdea A, Halder S, Krusienski D J. An Auditory Oddball (P300) Spelling System forBrain-Computer Interfaces. Psychophysiology,2009,46(3):617–625.
[16] Krusienski D J, Sellers E W., McFarland D J, et al. Towanrd Enhanced P300SpellerPerformance[J]. Journal of Neuroscience Methods,2008,167(1):15–21.
[17] Wang Y J, Gao S K, Gao X R. Common Spatial Pattern Method for Channel Selection in MotorImagery Based Brain-Computer Interface[C]. IEEE EMBS27th annual conference2005.Shanghai, China, Sept1-4,2005:5392–5395.
[18] Sun S L, Zhang C SH. An Optimal Kernel Feature Extractor and Its Application to EEG SignalClassification[J]. Neurocomputing2005,69(13–15):1743–1748.
[19] Krusienski D J, Sellers E W, Mcfarland D J, et al. Toward Enhanced P300Speller Performance[J].Journal of Neuroscience Methods,2008,167(1):15–21.
[20]张旭秀,陈坚.基于Fisher距离的新型脑机接口分类器[J].大连交通大学学报,2010,(02):104–107.
[21]唐艳,柳建新,龚安栋.脑-机接口中新的脑电数据分类方法[J].电子科技大学学报,2009,38(6):1034–1038.
[22] Nasihatkon B, Boostani R, Jahromi M. An Efficient Hybrid Linear and Kernel CSP Approachfor EEG Feature Extraction[J]. Neurocomputing.2009,73(1-3):432–437.
[23]刘广权,黄淦,朱向阳.共空域模式方法在多类别分类中的应用[J].中国生物医学工程学报,2009,28(6):935–938.
[24]李映,焦李成.基于核Fisher判别分析的目标识别[J].西安电子科技大学学报,2003,30(2):179–182.
[25] Baudat G, Anouar F. Generalized Discriminant Analysis Using a Kernel Approach[J], NeuralComput.,2000,12(10):2385–2404.
[26] Wang Y., Ruan Q. Kernel Fisher Discriminant Analysis for Palmprint Recognition[C]. IEEE The18th Int. Conf. on Pattern Recognition, ICPR’06,2006:457–460.
[27]赵海龙,穆志纯,张霞,基于LDA/GSVD和支持向量机的人耳识别[J].上海理工大学学报,2009,6(1):601–608.
[28] Wang Y J, Zhang Z G, Li Y, et al. BCI Competition2003—Data Set IV:An Algorithm Based onCSSD and FDA for Classifying Single-trial EEG[J]. IEEE Transactions on Bio-MedicalEngineering2004,51(6):1081–1086.
[29] Lotte F, Lécuyer A, Lamarche F, et al. Studying the Use of Fuzzy Inference Systems for MotorImagery Classification[J]. IEEE Transactions on Neural System and Rehabilitation Engineering,2007,15(2):322–324.
[30] Brodu N, Lotte F, Lécuyer A. Comparative study of band-power extraction techniques for motorimagery [C], IEEE Symposium Series on Computational Intelligence, Cognitive Algorithms,Mind, and Brain,2011:95–100.
[31]贺云辉,赵力,邹采荣.基于核鉴别共同矢量的小样本脸像鉴别方法[J];电子与信息学报,2006,28(12):2296–2300.
[32] Cevikalp H, Neamtu M, Wilkes M,et al.Discriminative Common Vectors for Face Recognition[J].Pattern Analysis and Machine Intelligence, IEEE Trans actions on,2005,27(1):4–13.
[33] Cevikalp H, Neamtu M, Wilkes M.Discriminative Common Vectors Method with Kernels[J],IEEE Trans. Neural Netw.,2006,17(6):1550–1565.
[34] Logar V, Belic A.Brain–Computer Interface Analysis of a Dynamic Visuo-Motor Task[J]. ArtificialIntelligence in Medicine,2011,51(1):43–51.
[35] Dean J K, Eric W S, Sabri B, et al. A Comparison of Classification Techniques for the P300Speller[J]. Journal of Neural Engineering.2006,3(4):299–305.
[36] Townsend G, Graimann B, Pfurtscheller G. Continuous EEG Classification During MotorImagery-Simulation of an Asynchronous BCI [J]. IEEE Transactions on Neural Systems andRehabilitation Engineering,2004,12(2):258–265.
[37] Martens S, Hill N, Farquhar J, et al. Overlap and Refractory Effects in a Brain–Computer InterfaceSpeller Based on the Visual P300Event-Related Potential[J]. Journal of Neural Engineering,2009,6(2):026003–026003.
[38] Allison B, McFarland D, Schalk G, et al. Towards an Independent Brain-Computer Interface UsingSteady State Visual Evoked Potentials[J].Clinical neurophysiology,2008,119(2):399–408.
[39] Kaper M, Meinicke P, Grossekathoefer U, et al. BCI Competition2003—Data Set IIb: SupportVector Machines for the P300Speller Paradigm[J]. IEEE Transactions on BiomedicalEngineering,2004,51(6):1073–1076.
[40]卢文喜，李俊，于福荣,等.逐步判别分析法在筛选水质评价因子中的应用[J].吉林大学学报.2009,39(1):126–130.
[41] Sellers E W, Donchin E. A P300-based brain-computer-interface: Initial tests by ALS patients[J].Clinical Neurophysiology,2006,117(3):538–548.
[42]赵丽,万柏坤.基于P300脑-机接口系统研究[J].天津工程师范学院学报,2005,15(2):5–9.
[43] Sellers E W, Kubler A, Donchin E. Brain-computer interface research at the university of southFlorida cognitive psychophysiology laboratory: the P300speller[J]. IEEE Transaction on NeuralSystems and Rehabilitation Engineering,2006,14(2):221–228.
[44]官金安,陈亚光.相干平均单次提取脑-机接口信号[J].华中科技大学学报,2007,35(1):11–13.
[45] Lin Z L, Zhang C S.Enhancing classification by perceptual characteristic for the P300spellerparadigm[C].Proceedings of the2nd International IEEE EMBS Conference on NeuralEngineering.IEEE,2005:574–576.
[46] Liu Y, Zhou Z T, Hu D W, et al.T-weighted approach for neural information processing in P300based brain-computer interfaces[C].2005International Conference on Neural Networks and Brain.2005,4(4):1535–1539.
[47]赵向东.视觉事件相关电位[J].现代电生理学杂志,2003,10(3):26–30.
[48]马建军,孙相如,吴逊.事件相关电位的研究进展[J].国外医学神经病学神经外科学分册,1994,21(1):134–138.
[49]杨文俊.事件相关电位简述[J].中华神经精神科杂志,1991,24(5):88–93.
[50]罗越嘉.事件相关电位中的P300成分[J].物理医学与康复学分册,1990,10(3):22–26.
[2]王毅军.基于节律调制的脑－机接口系统—从离线到在线的跨越[D].北京：清华大学生物医学工程学科博士学位论文，2007:14–16.
[3] Schalk G, McFarland D J, Hinterberger T,et al. BCI2000:A General-Purpose Brain-ComputerInterface(BCI)System[C]. IEEE Trans Biomed Eng.2004,51(06):1034–1043.
[4] Schalk G, Jürgen M. A Pratical Guide to Brain-Computer Interfacing with BCI2000[M]. SpringerLondon Dordrecht Heidelberg New York.2010:4–5.
[5]王金甲,周丽娜,赵玉超.基于MEG的脑机接口特征提取方法研究[J].仪器仪表学报,2010,31(7):1460–1465.
[6] Birch G E, Mason S G, Borisoff J F. Current Trends in Brain-Computer Interface Research at theNeil Squire Foundation[J]. IEEE Transaction on Neural Systems and Rehabilitation Engineering,2003,11(2):123–126.
[7] Pfurtscheller G, Neuper C. Graze-BCI:State of the Art and Clinical Applications[J]. IEEETransaction on Neural Systems and Rehabilitation Engineering,2003,11(2):177–180.
[8] Gao X R, Xu D F, Cheng M. A BCI-based Environmental Controller for the Motiondisabled[J].IEEE Transaction on Neural Systems and Rehabilitation Engineering,2003,11(2):137–140.
[9]边肇祺,张学工.模式识别[M].北京:清华大学出版社,2000:84–90.
[10]马忠伟,高上凯.基于P300的脑-机接口：视觉刺激强度对性能的影响[J].清华大学学报,2008,48(3):415–418.
[11] Hoffmann U, Vesin J M, Ebrahimi T, et al. An Efficient P300-based Brain-Computer Interface forDisabled Subjects[J]. Journal of Neuroscience Methods.2008,167(1):115–125.
[12] Finke A, Lenhardt A, Ritter H. The MindGame: A P300-based Brain-Computer Interface Game[J].Neural Networks.2009,22(9):1329–1333.
[13] Zhang H, Guan C, Wang C. Asynchronous P300-Based Brain--Computer Interfaces: AComputational Approach With Statistical Models[J]. Biomedical Engineering, IEEE Transactionson,2008,55(6):1754–176.
[14]吴边,苏煜,张剑慧,等.基于P300电位的新型BCI中文输入虚拟键盘系统[J].电子学报,2009,37(8):1733–1738.
[15] Furdea A, Halder S, Krusienski D J. An Auditory Oddball (P300) Spelling System forBrain-Computer Interfaces. Psychophysiology,2009,46(3):617–625.
[16] Krusienski D J, Sellers E W., McFarland D J, et al. Towanrd Enhanced P300SpellerPerformance[J]. Journal of Neuroscience Methods,2008,167(1):15–21.
[17] Wang Y J, Gao S K, Gao X R. Common Spatial Pattern Method for Channel Selection in MotorImagery Based Brain-Computer Interface[C]. IEEE EMBS27th annual conference2005.Shanghai, China, Sept1-4,2005:5392–5395.
[18] Sun S L, Zhang C SH. An Optimal Kernel Feature Extractor and Its Application to EEG SignalClassification[J]. Neurocomputing2005,69(13–15):1743–1748.
[19] Krusienski D J, Sellers E W, Mcfarland D J, et al. Toward Enhanced P300Speller Performance[J].Journal of Neuroscience Methods,2008,167(1):15–21.
[20]张旭秀,陈坚.基于Fisher距离的新型脑机接口分类器[J].大连交通大学学报,2010,(02):104–107.
[21]唐艳,柳建新,龚安栋.脑-机接口中新的脑电数据分类方法[J].电子科技大学学报,2009,38(6):1034–1038.
[22] Nasihatkon B, Boostani R, Jahromi M. An Efficient Hybrid Linear and Kernel CSP Approachfor EEG Feature Extraction[J]. Neurocomputing.2009,73(1-3):432–437.
[23]刘广权,黄淦,朱向阳.共空域模式方法在多类别分类中的应用[J].中国生物医学工程学报,2009,28(6):935–938.
[24]李映,焦李成.基于核Fisher判别分析的目标识别[J].西安电子科技大学学报,2003,30(2):179–182.
[25] Baudat G, Anouar F. Generalized Discriminant Analysis Using a Kernel Approach[J], NeuralComput.,2000,12(10):2385–2404.
[26] Wang Y., Ruan Q. Kernel Fisher Discriminant Analysis for Palmprint Recognition[C]. IEEE The18th Int. Conf. on Pattern Recognition, ICPR’06,2006:457–460.
[27]赵海龙,穆志纯,张霞,基于LDA/GSVD和支持向量机的人耳识别[J].上海理工大学学报,2009,6(1):601–608.
[28] Wang Y J, Zhang Z G, Li Y, et al. BCI Competition2003—Data Set IV:An Algorithm Based onCSSD and FDA for Classifying Single-trial EEG[J]. IEEE Transactions on Bio-MedicalEngineering2004,51(6):1081–1086.
[29] Lotte F, Lécuyer A, Lamarche F, et al. Studying the Use of Fuzzy Inference Systems for MotorImagery Classification[J]. IEEE Transactions on Neural System and Rehabilitation Engineering,2007,15(2):322–324.
[30] Brodu N, Lotte F, Lécuyer A. Comparative study of band-power extraction techniques for motorimagery [C], IEEE Symposium Series on Computational Intelligence, Cognitive Algorithms,Mind, and Brain,2011:95–100.
[31]贺云辉,赵力,邹采荣.基于核鉴别共同矢量的小样本脸像鉴别方法[J];电子与信息学报,2006,28(12):2296–2300.
[32] Cevikalp H, Neamtu M, Wilkes M,et al.Discriminative Common Vectors for Face Recognition[J].Pattern Analysis and Machine Intelligence, IEEE Trans actions on,2005,27(1):4–13.
[33] Cevikalp H, Neamtu M, Wilkes M.Discriminative Common Vectors Method with Kernels[J],IEEE Trans. Neural Netw.,2006,17(6):1550–1565.
[34] Logar V, Belic A.Brain–Computer Interface Analysis of a Dynamic Visuo-Motor Task[J]. ArtificialIntelligence in Medicine,2011,51(1):43–51.
[35] Dean J K, Eric W S, Sabri B, et al. A Comparison of Classification Techniques for the P300Speller[J]. Journal of Neural Engineering.2006,3(4):299–305.
[36] Townsend G, Graimann B, Pfurtscheller G. Continuous EEG Classification During MotorImagery-Simulation of an Asynchronous BCI [J]. IEEE Transactions on Neural Systems andRehabilitation Engineering,2004,12(2):258–265.
[37] Martens S, Hill N, Farquhar J, et al. Overlap and Refractory Effects in a Brain–Computer InterfaceSpeller Based on the Visual P300Event-Related Potential[J]. Journal of Neural Engineering,2009,6(2):026003–026003.
[38] Allison B, McFarland D, Schalk G, et al. Towards an Independent Brain-Computer Interface UsingSteady State Visual Evoked Potentials[J].Clinical neurophysiology,2008,119(2):399–408.
[39] Kaper M, Meinicke P, Grossekathoefer U, et al. BCI Competition2003—Data Set IIb: SupportVector Machines for the P300Speller Paradigm[J]. IEEE Transactions on BiomedicalEngineering,2004,51(6):1073–1076.
[40]卢文喜，李俊，于福荣,等.逐步判别分析法在筛选水质评价因子中的应用[J].吉林大学学报.2009,39(1):126–130.
[41] Sellers E W, Donchin E. A P300-based brain-computer-interface: Initial tests by ALS patients[J].Clinical Neurophysiology,2006,117(3):538–548.
[42]赵丽,万柏坤.基于P300脑-机接口系统研究[J].天津工程师范学院学报,2005,15(2):5–9.
[43] Sellers E W, Kubler A, Donchin E. Brain-computer interface research at the university of southFlorida cognitive psychophysiology laboratory: the P300speller[J]. IEEE Transaction on NeuralSystems and Rehabilitation Engineering,2006,14(2):221–228.
[44]官金安,陈亚光.相干平均单次提取脑-机接口信号[J].华中科技大学学报,2007,35(1):11–13.
[45] Lin Z L, Zhang C S.Enhancing classification by perceptual characteristic for the P300spellerparadigm[C].Proceedings of the2nd International IEEE EMBS Conference on NeuralEngineering.IEEE,2005:574–576.
[46] Liu Y, Zhou Z T, Hu D W, et al.T-weighted approach for neural information processing in P300based brain-computer interfaces[C].2005International Conference on Neural Networks and Brain.2005,4(4):1535–1539.
[47]赵向东.视觉事件相关电位[J].现代电生理学杂志,2003,10(3):26–30.
[48]马建军,孙相如,吴逊.事件相关电位的研究进展[J].国外医学神经病学神经外科学分册,1994,21(1):134–138.
[49]杨文俊.事件相关电位简述[J].中华神经精神科杂志,1991,24(5):88–93.
[50]罗越嘉.事件相关电位中的P300成分[J].物理医学与康复学分册,1990,10(3):22–26.