在睡眠障礙的診斷中多頻道腦波量測設備被視為一黃金指標，然而其在使用上較為不便、價格昂貴且不適合作為居家量測設備，而單頻道腦波量測裝置解決了多頻道腦波設備大部分的問題，但是單頻道腦波量測裝置在訊號分析上因頻道較少而有所限制，因此我們必須去蕪存菁，在單頻道訊號中萃取出重要資訊。
由於失眠為最常見的睡眠障礙，因此本文旨在開發以單頻道額前訊號區分正常人與失眠患者的演算法。藉由三種不同的分析方式，找出受測者於睡眠時單頻道前額訊號的特徵指標，並以這些特徵指標作為區分正常人與失眠患者的依據，藉此開發出一有效區分兩族群的分類演算法，而三個分析方法分別為：(1)以訊號的改變速度進行分析，使用訊號的越零點間距作為分析訊號改變快慢的分析方法，間距越大代表訊號改變較慢，反之間距小則代表改變得快、(2)訊號的時頻特性，以連續小波轉換分析訊號並得到時頻圖，以此觀察訊號的時頻特性、(3)訊號的複雜度，以多尺度熵來分析越零點間距及各頻帶訊號改變的複雜度。而除了以上述三種方法直接分析訊號外，我們會將各分析方法得到的結果個別進行大小排序計算PR值，以去除因受測者睡眠長度不同所造成時間上的影響，接著將各個分析方法排序的結果進行統計學Wilcoxon's rank-sum test找出有顯著差異的特徵作為特徵指標，最後則將這些特徵指標兩兩搭配以二次判別分析(QDA)進行分類得到分類結果。
在各種分析方式中所得到的特徵指標共有23種，將這23種特徵指標兩兩搭配進行分類得到的最佳分類結果為以越零點間距指標搭配小波轉換指標所得到的結果，其精準度達到97.8%，此外總共有50種分類結果的精準度超過90%。
從以上的分類結果得知本文於單頻道前額腦波訊號的分析方法能夠有效的區分正常人族群與失眠患者族群，搭配上單頻道量測裝置得方便性，非常適合用於居家醫療。因此，我們所提出的演算法可以作為初步的居家失眠篩選工具。

Polysomnography (PSG) is considered the gold standard in the diagnosis of sleep disordered breathing. However, compared with the single channel EEG sensor, PSG is difficult to operate, expensive and not suitable for home use. However, with fewer number of signals, the single channel senor provides much less information than PSG. Hence, extracting sufficient information from a single channel signal is a very important task.
Considering the fact that insomnia is the most prevalent sleep disorder, this thesis is aimed to develop algorithms to distinguish insomnia patients form normal people via single channel forehead signal. In specific, we use three different approaches to analyze the forehead signal during sleep in order to identify features that have significantly different values between insomnia patients and normal people. The first method uses the zero crossing point interval to characterize the speed of the signal. The second method uses the continuous wavelet transform (CWT) to generate a series of features to assess the time-frequency domain properties of the signal. The third method uses the multiscale entropy criterion to quantify the complexity of the signal. In addition to comparing the value of the features, we also used the percentile rank of these features. By using the Wilcoxon's rank-sum test to find out which of these features have significantly different median values in the two groups, 23 features were selected to classify insomnia patients from normal people by using Quadratic Discriminant Analysis (QDA).
Selected from these 23 features, we tested many pairs of features. The best classification result was obtained by the combination of the zero crossing point interval and a CWT feature. The best classification accuracy is 97.8%. Additionally, we have identified 50 different feature pairs that can achieve classification accuracy of 90% or higher.
According to these promising classification results, our algorithm has the potential to effectively distinguish insomnia patients from normal people. With the simplicity of single channel measurement requirement, the proposed approach is ideal for home use. As a result, the proposed method has the potential to become a home-based insomnia screening tool.

論文審定書 i
論文公開授權書 ii
謝誌 iii
摘要 iv
Abstract v
目錄 vii
圖次 x
表次 xvi
第一章 緒論 1
1.1 前言 1
1.2 研究動機與目的 2
1.3 文獻回顧 3
第二章 實驗方法 4
2.1 Relax單頻道額前訊號量測套件 4
2.1.1 Relax與其他腦波量測設備比較[8][9][10][11] 4
2.1.2 Relax套件之硬體規格 5
2.1.3實際量測結果 7
2.2 收案流程 11
2.3 受測對象 12
2.4 Relax套件與血氧濃度計的量測方式 13
2.4.1 Relax套件的操作方式 13
2.4.2血氧濃度計配戴方式 15
2.4.3睡眠資料紀錄表 15
2.5 Relax睡眠訊號篩選 16
2.6 以血氧濃度計篩選訊號 18
第三章 分析方法與各項指標 19
3.1 分析流程 19
3.2 訊號的前處理 20
3.3 越零點與越零點間距 22
3.3.1 越零點(Zero Crossing Point) 22
3.3.2越零點間距(Zero Crossing Points Interval) 23
3.4 多尺度熵(Multi-Scale Entropy, MSE) 24
3.4.1多尺度熵的計算方法 24
3.4.2多尺度熵及經Shuffle後的多尺度熵 26
3.4.3多尺度熵隨時間的變化 28
3.5 連續小波轉換(Continuous Wavelet Transform) 29
3.5.1 Morlet小波函數[25] 29
3.5.2實數Morlet小波函數的限制 31
3.5.3複數Morlet小波函數[26] 32
3.5.4小波轉換於前額訊號的分析 35
3.6 以PR值來比較各睡眠特徵 37
3.7 判別分析(Discriminant Analysis) 40
第四章 結果與討論 41
4.1 越零點間距分析睡眠訊號結果 42
4.1.1 前兩小時睡眠訊號比較結果 42
4.1.2 整夜睡眠訊號比較結果 47
4.2 越零點間距的多尺度熵值 51
4.3 連續小波轉換分析睡眠訊號的結果 62
4.3.1 前兩小時睡眠訊號比較結果 62
4.3.2 整夜睡眠訊號比較結果 70
4.4 連續小波轉換的多尺度熵 77
4.5 以二次判別分析分類結果 86
4.5.1 最佳的分類結果 87
4.5.2 各指標的分類效果 88
4.5.3 各類別指標的分類效果 90
4.5.4 以委員會機器演算法進行分類 91
第五章 結論 93
參考文獻 97
附錄 101

[1] Riemann D., Berger M., Voderholzer U., Sleep and depression — results from psychobiological studies: an overview. Biol Psychol. 2001. Jul-Aug; 57(1-3):67-103.
[2] Bonnet M.H., Arand D.L.., Hyperarousal and insomnia: state of the science. Sleep Med Rev. 2010 Feb; 14(1):17.
[3] Freedman R.R., EEG power spectra in sleep-onset insomnia. Electroencephalogr Clin Neurophysiol. 1986 May; 63(5):408-13.
[4] Lamarche C.H., Ogilvie R.D., Electrophysiological changes during the sleep onset period of psychophysiological insomniacs, psychiatric insomniacs, and normal sleepers. Sleep. 1997 Sep; 20(9):724-33.
[5] Merica H., Blois R., Gaillard J.M., Spectral characteristics of sleep EEG in chronic insomnia. Eur J Neurosci. 1998 May; 10(5):1826-34.
[6] Merica H., Gaillard J.M., The EEG of the sleep onset period in insomnia: a discriminant analysis. Physiol Behav. 1992 Aug; 52(2):199-204.
[7] Lee Y.Y., Chen Y.S., Chen L.F., Automated Sleep Staging Using Single EEG Channel for REM Sleep Deprivation. IEEE. 2009 June; 439 - 42.
[8] Fraiwan L.A., Khaswaneh N.Y., Lweesy K.Y., Automatic Sleep Stage Scoring with Wavelet Packets Based on Single EEG Recording. World Academy of Science, Engineering and Technology. 2009; 54(3):485-88
[9] Kuo T.B., Chen C.Y., Hsu Y.C., Yang C.C., Performance of the frequency domain indices with respect to sleep staging. Clin Neurophysiol. 2012 Jul; 123(7):1338-45.
[10] Achermann P., EEG Analysis Applied to Sleep. Epileptologie 2009; 26: 28 – 33

[11] Cahn B.R., Polich J., Meditation states and traits: EEG, ERP, and neuroimaging studies. Psychol Bull. 2006 Mar; 132(2):180-211.
[12] Gyulay S., Olson L.G., Hensley M.J., King M.T., Allen K.M., Saunders N.A., A comparison of clinical assessment and home oximetry in the diagnosis of obstructive sleep apnea. Am Rev Respir Dis. 1993 Jan; 147(1):50-3.
[13] Vazquez J., Tsai W., Flemons W., Masuda A., Brant R., Hajduk E., Whitelaw W., Remmers J., Automated analysis of digital oximetry in the diagnosis of obstructive sleep apnoea. Electroencephalogr Clin Neurophysiol. 1986 May; 63(5):408-13.
[14] Golpe R., Jiménez A., Carpizo R., Cifrian J.M., Utility of home oximetry as a screening test for patients with moderate to severe symptoms of obstructive sleep apnea. Sleep. 1999 Nov 1; 22(7):932-7.
[15] ERTL J.P., Detection of evoked potentials by zero crossing analysis. Electroencephalogr Clin Neurophysiol. 1965 May; 18:630-1.
[16] Shahidi Z. A., Tafreshi R., Javidan M., Dumont G.A.., Predicting temporal lobe epileptic seizures based on zero-crossing interval analysis in scalp EEG. Conf Proc IEEE Eng Med Biol Soc. 2010; 2010:5537-40.
[17] Costa M., Goldberger A.L., Peng C.K., Multiscale Entropy of human gait dynamics. Physica A: Statistical Mechanics and its Applications, 2003 Dec 1, Volume 330, Issues 1–2, 53–60.
[18] Costa M., Goldberger A.L., Peng C.K., Multiscale entropy analysis of biological signals. Phys Rev E Stat Nonlin Soft Matter Phys. 2005 Feb; 71(2 Pt 1):021906.
[19] Costa M., Goldberger A.L., Peng C.K., Multiscale entropy analysis of complex physiologic time series. Phys Rev Lett. 2002 Aug 5; 89(6):068102.

[20] Okazaki R., Takahashi T., Ueno K., Takahashi K., Ishitobi M., Kikuchi M., Higashima M., Wada Y., Changes in EEG complexity with electroconvulsive therapy in a patient with autism spectrum disorders: a multiscale entropy approach. Front Hum Neurosci. 2015 Feb 26; 9:106.
[21] Bell I.R., Howerter A., Jackson N., Aickin M., Bootzin R.R., Nonlinear dynamical systems effects of homeopathic remedies on multiscale entropy and correlation dimension of slow wave sleep EEG in young adults with histories of coffee-induced insomnia. Homeopathy. 2012 Jul; 101(3):182-92.
[22] Lu C.F., Teng S., Hung C.I., Tseng P.J., Lin L.T., Lee P.L., Wu Y.T., Reorganization of functional connectivity during the motor task using EEG time–frequency cross mutual information analysis. Clin Neurophysiol. 2011 Aug; 122(8):1569-79.
[23] Vialatte F.B., Martin C., Dubois R., Haddad J., A machine learning approach to the analysis of time-frequency maps, and its application to neural dynamics. Neural Netw. 2007 Mar; 20(2):194-209.
[24] Semmlow J.L., Biosignal and Biomedical Image Processing: MATLAB-Based pplications. Marcel Dekker, Inc., 2004, 177-210.
[25] Cohen M.X., Analyzing Neural Time Series Data: Theory and Practice. Issues in Clinical and Cognitive Neuropsychology, 2014, Chapter 13, 141-150.
[26] Cohen M.X., Analyzing Neural Time Series Data: Theory and Practice. Issues in Clinical and Cognitive Neuropsychology, 2014, Chapter 14, 151-174.
[27] Fisher, R.A., The Use of Multiple Measurements in Taxonomic Problems. Annals of Eugenics, Vol. 7, pp. 179–188,
[28] Krzanowski W.J., Principles of Multivariate Analysis: A User's Perspective. New York: Oxford University Press, 1988.

[29] Seber G.A.F., Multivariate Observations. Hoboken, NJ: John Wiley & Sons, Inc., 1984.
[30] Santosh S., Maya R.G., Bela A.F., Bayesian Quadratic Discriminant Analysis. Journal of Machine Learning Research 8, 2007; 1277-1305.
[31] Villarroya A., Ríos M., Oller J.M., Discriminant analysis algorithm based on a distance function and on a Bayesian decision. Biometrics. 1995 Sep;51(3):908-19.