為了建立一套適用於即時腦波辨識分析系統的分析流程，正確的從單筆腦電波訊號中擷取出有利於事件判別的特徵是相當重要的。於本研究中，我們運用交互信息法(cross mutual information, CMI)來發展左右手臂上抬運動之判斷流程並討論於流程中的兩層分類方法之效用。
我們針對不同受試者利用時間頻率圖形尋找其各自適合的頻帶，在第一層上抬運動與休息狀態分類上，我們利用傳統β能量反彈(β rebound)特徵與交互信息特徵比較；而第二層左手臂和右手臂上抬運動分類，我們取手臂運動試驗中特徵顯著的時間區段訊號，接著以三種具不同特性之濾波方式：Morlet小波轉換(Morlet wavelet transform)濾波、本質模態函數(intrinsic mode function， IMF)濾波、巴氏帶通濾波(Butterworth band pass filter)帶入交互信息法，最後比較不同濾波方式與不同交互信息特徵組合於辨識之效果。
實驗結果顯示，在上抬運動與休息狀態的分類上，交互信息特徵有與β能量一樣的辨識能力，但比較訓練資料與測試資料的差異後，發現交互信息特徵有資料穩定度差的缺點。而左右手辨識結果，巴氏帶通濾波佐以全腦交互信息組成圖平均辨識率達83%±13%，是不同的比較中最佳的組合。另外，根據Mann-Whitney U檢定發現帶入頻帶是否為受試者與事件相關的頻帶之辨識率在統計上具有顯著差異(p≤0.014)。而帶入有顯著特徵的的時間區段訊號也較帶入完整的時間區段訊號對於不同特徵擷取整體平均獲得高7%的辨識率。於是我們建議運用受試者與事件相關的頻帶訊號於分析流程上，利用β能量偵測動作狀態，並使用巴氏帶通濾波於有顯著特徵的時間區段訊號佐以全腦交互信息組成圖來辨識左右手運動。

This thesis aims to develop an algorithm composed of two-level classifiers based on single-trial electroencephalography (EEG) signal for the recognition of movement versus resting state and subsequently left arm movement versus right arm movement. The first step of algorithm is feature extraction. We estimate a subject-specific frequency band related to the arm movement from time frequency map of each subject. The first-level classifier is designed to classify the movement versus resting state in which either the cross mutual information (CMI) or the conventional beta-rebound power feature are employed. The second-level classifier is designed to recognize the left- or right-arm movement where the significant temporal period is defined followed by one of the three filters for the feature extractions: Morlet wavelet transform filters, intrinsic mode function (IMF) filters, Butterworth band pass filters. Once EEG signals on all channels were processed by these filters, the CMIs of filtered signals between any two channels were computed, respectively.
The recognition results of motion versus resting state show that the recognition rates from the CMI and beta-rebound power feature are comparable. In the second-level classification, the averaged classification rates of left- versus right-arm movement obtained from the Butterworth band pass filter with whole-brain CMI maps are 83%±13% which are superior to that from Morlet wavelet transform and IMF. In addition, the recognition rates resulted from subject-specific frequency bands have a significant difference ( p≤0.014, Mann-Whitney U method) compared with other bands. Moreover, the recognition rate obtained from using the significant temporal period is 7% higher than that from using the whole epoch. In conclusion, we suggest using the beta-rebound power as a feature for the detection of motion versus resting state and using the Butterworth band pass filter with whole-brain CMI maps on signal of significant temporal period for the recognition of left- versus right-arm movement.

致謝 i
中文摘要 iii
Abstract v
目錄 vi
圖表索引 viii
第一章緒論 1
1.1 前言 1
1.2 認識腦波 3
1.2.1 腦波的頻率 4
1.2.2 腦波律動的功能區域性 5
1.3 事件相關之同步/非同步腦波 6
1.3.1 事件相關的電位 6
1.3.2 事件相關之非同步腦波與事件相關之同步腦波 7
1.3.3 事件相關之非同步腦波與事件相關之同步腦波的分析方法 8
1.4 腦波訊號之擷取 10
1.4.1 腦電波儀 10
1.4.2 頻道組合之模式 10
1.4.3 腦波電極配置法 11
1.4.4紀錄腦電波的影響因素 13
1.5 研究的動機 14
第二章 實驗材料與分析方法 16
2.1實驗材料 16
2.1.1 受試者 16
2.1.2 硬體設備-腦電波儀 17
2.1.3 實驗設計 18
2.2 資料分析演算法 20
2.2.1 交互信息法 22
2.2.2 決定受試者的頻帶 25
2.2.3 上抬運動與休息狀態之分類 27
2.2.4 左手臂和右手臂上抬運動之分類 31
2.2.4.1時間區段的選擇 36
2.2.4.2 Morlet小波轉換 39
2.2.4.3經驗模態分解法 42
2.2.5分類的方法 45
2.2.5.1主要成份分析法 45
2.2.5.2費雪線性區別分析法 48
第三章 結果 52
3.1 交互信息特徵組合 53
3.2 上抬運動與休息狀態之分類 55
3.2.1 決定ICP範圍及訓練資料 55
3.2.2 正確率 57
3.2.3 β能量和交互信息的特徵 59
3.2.4 ROC曲線與利用訓練資料選定閾值 64
3.3 左手臂和右手臂上抬運動之分類 67
3.3.1 時間區段的選擇 67
3.3.2 正確率 68
3.3.3 左右手特徵樣式的差異 69
第四章 討論與結論 73
4.1 討論 73
4.1.1 交互信息法取樣數(sampling bins)的選擇 73
4.1.2 選擇特徵顯著時間區段之影響 76
4.1.3 選擇受試者與事件相關頻帶之影響 77
4.1.4 判斷流程的選擇 80
4.2 結論 83
參考文獻 85

1. 吳進安等編著， 神經診斷學， 國立編譯館， 1998年3月。
2. 林景福等編著， 圖解腦波入門， 1985年4月。
3. Fisch BJ. Fisch and Spehlmann's EEG Primer.-3rd ed: Elsevier; 1999.
4. Pfurtscheller G. ERD as an index of anticipatory behavior. Handbook of electroencephalography and clinical neurophysiology revised series. Volume 6; 1999. p 203-217.
5. Lopes da Silva FH, Pfurtscheller G. Basic concepts on EEG synchronization and desynchronization. Handbook of electroencephalography and clinical neurophysiology revised series. Volume 6; 1999. p 203-217.
6. Pfurtscheller G, Lopes da Silva FH. Event-related EEG/MEG synchronization and desynchronization: basic principles. Clin Neurophysiol 1999;110(11):1842-1857.
7. Pfurtscheller G, Aranibar A. Event-related cortical desynchronization detected by power measurements of scalp EEG. Electroencephalogr Clin Neurophysiol 1977;42(6):817-826.
8. Pfurtscheller G. Quantification of ERD and ERS in the time domain. Handbook of electroencephalography and clinical neurophysiology. Volume 6; 1999. p 89-105.
9. Pfurtscheller G, Aranibar A. Evaluation of event-related desynchronization (ERD) preceding and following voluntary self-paced movement. Electroencephalogr Clin Neurophysiol 1979;46(2):138-146.
10. Salmelin R, Forss N, Knuutila J, Hari R. Bilateral activation of the human somatomotor cortex by distal hand movements. Electroencephalogr Clin Neurophysiol 1995;95(6):444-452.
11. Clochon P, Fontbonne J, Lebrun N, Etevenon P. A new method for quantifying EEG event-related desynchronization:amplitude envelope analysis. Electroencephalogr Clin Neurophysiol 1996;98(2):126-129.
12. Jasper H. The 10-20 electrode system of the International Federation. Electroencephalography and Clinical Neurophysiology 1958;10:371-375.
13. Malmivuo J, Plonsey R. Bioelectromagnetism. Principles and applications of bioelectric and biomagnetic fields: Oxford University Press; 1995.
14. 關尚勇、林吉合編著， 破解腦電波， 2002年。
15. Friston KJ. Transients, metastability, and neuronal dynamics. Neuroimage 1997;5(2):164-171.
16. Tononi G SO, Edelman GM. A measure for brain complexity: relating functional segregation and integration in the nervous system. Proc Natl Acad Sci U S A 1994;91(11):5033-5037.
17. Fraser AM, Swinney HL. Independent coordinates for strange attractors from mutual information. Phys Rev A 1986;33(2):1134-1140.
18. Jeong J, Gore JC, Peterson BS. Mutual information analysis of the EEG in patients with Alzheimer's disease. Clin Neurophysiol 2001;112(5):827-835.
19. David O, Cosmelli D, Friston KJ. Evaluation of different measures of functional connectivity using a neural mass model. Neuroimage 2004;21(2):659-673.
20. Shannon CE. A mathematical theory of communication. Bell System Technical Journal 1948;27:379-423/623-656.
21. Pluim JP, Maintz JB, Viergever MA. Mutual-information-based registration of medical images: a survey. IEEE Trans Med Imaging 2003;22(8):986-1004.
22. Chen CC, Hsieh JC, Wu YZ, Lee PL, Chen SS, Niddam DM, Yeh TC, Wu YT. Mutual-information-based approach for neural connectivity during self-paced finger lifting task. Hum Brain Mapp 2008;29(3):265-280.
23. Pfurtscheller G, Solis-Escalante T. Could the beta rebound in the EEG be suitable to realize a "brain switch"? Clin Neurophysiol 2009;120(1):24-29.
24. Toro R, Fox PT, Paus T. Functional coactivation map of the human brain. Cereb Cortex 2008;18(11):2553-2559.
25. Oppenheim AV, Schafer RW. Discrete-time Signal Processing. Prentice-Hall International I, editor; 1989.
26. Sakkalis V, Zervakis M. Methodological Framework for EEG Feature Selection Based on Spectral and Temporal Profiles. In: CHAOVALITWONGSE W, editor. Computational Neuroscience. Volume 38; 2010.
27. Sakkalis V, Zervakis M, Micheloyannis S. Significant EEG features involved in mathematical reasoning: evidence from wavelet analysis. Brain Topogr 2006;19(1-2):53-60.
28. Meyer Y. Wavelets: Algorithms and Applications: Siam; 1993.
29. Huang NE, Shen Z, Long SR, Wu MC, Shih HH, Zheng Q, Yen NC, Tung CC, Liu HH. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proc R Soc Lond A 1998;454:903-995.
30. Huang M, Wu P, Liu Y, Bi L, Chen H. Application and Contrast in Brain-Computer Interface between Hilbert-Huang Transform and Wavelet Transform. The 9th International Conference for Young Computer Scientists 2008:1706-1710.
31. Wang L, Xu G, Wang J, Yang S, Yan W. Application of Hilbert-Huang transform for the study of motor imagery tasks. Conf Proc IEEE Eng Med Biol Soc 2008;2008:3848-3851.
32. Duda RO, Hart PE, Stork DG. Pattern Classification, 2nd Ed: John Wiley &Sons, Inc.; 2001.
33. Yang K, Shahabi C. On the stationarity of multivariate time series for correlation-based data analysis. Proc of 5fth IEEE International Conference on Data Mining. Houston: Texas; 2005.
34. Zhang D, Zhou ZH, Chen S. Diagonal principal component analysis for face recognition. Pattern Recognition 2006;39:140-142.
35. Müller KR, Krauledat M, Dornhege G. Machine Learning Techniques for Brain-Computer Interfaces. Biomedical Engineering 2004.
36. Pfurtscheller G, Stancak A, Jr., Neuper C. Post-movement beta synchronization. A correlate of an idling motor area? Electroencephalogr Clin Neurophysiol 1996;98(4):281-293.
37. Lee PL, Wu YT, Chen LF, Chen YS, Cheng CM, Yeh TC, Ho LT, Chang MS, Hsieh JC. ICA-based spatiotemporal approach for single-trial analysis of postmovement MEG beta synchronization. Neuroimage 2003;20(4):2010-2030.
38. 陳建勝、呂兆文、陳美菁、朱瑞淵和呂明哲編著， 統計學商業與管理的應用， 2008年2月。