對於長時間的腦電波紀錄，如睡眠腦電波量測，由於醫事人員或是專業醫師需要消耗大量的時間與心力去判讀出腦電波的特徵，甚至腦電波病徵，因此造成判讀睡眠腦電波訊號的不便。有鑑於此，若能自動地對睡眠腦電波及睡眠狀態進行判讀及分類，將能大幅的減少時間上的成本及簡化各種繁雜的工作。本論文旨在利用適應性共振理論自我組織類神經網路對睡眠期進行自動分類的研究，受測者來至台中中港澄清綜合醫院，共23位無睡眠障礙的正常人(13男，10女，年齡: 21.47±1.53)。在類神經網路輸入樣本方面，本研究的取樣頻率為250Hz，每位受測者截取單一通道(C3-A1)約一小時的腦電波訊號，再取其中每段時距為連續三秒共750點並經漢寧視窗的資料做快速傅立葉轉換，並利用其13個頻帶的功率頻譜密度作為自我組織類神經網路輸入的特徵參數。在類神經網路訓練過程中，利用16位受測者睡眠狀態資料樣本作訓練，另外7位受測者睡眠狀態資料樣本作測試。結果對於清醒狀態的辨識正確率可達到98%，睡眠第一期的辨識率較不理想，睡眠第二期的辨識率可達到80.6%，對於睡眠慢波其辨識率可以達到97.5%。此實驗結果驗證了利用適應性共振理論類神經網路作自動辨識判讀睡眠腦電波訊號是一件可行的事，未來將透過樣本數的增加，並增加資料庫內之睡眠狀態模式預期可再提升辨識率。

The commonly used method to a sleep specialist for sleep stage classification is to apply visual inspection. This is a very time consuming and laborious task. Automatic sleep stage classification becomes necessary to facilitate this visual inspection process. In this study, a self-organizing neural network-adaptive resonance theory (ART) is presented to classify the sleep stage. To this aim, 23 subjects are recruited from Cheng-Ching General Hospital, Taichung, Taiwan, for this experiment. Among them are healthy adults (13 men, 10 women) with a mean age of 21.47±1.53 years old. The sampling frequency of electroencephalogram (EEG) data acquisition was set to be 250 Hz, and the one-hour single channel (C3-A1) EEG signal for each subject was acquired. Each epoch with consecutive 3-second segment was extracted using Hanning window. Followed with 13 frequency features based on power spectrum density of each epoch segment were calculated as the input of the ART neural network; In the training process, the sleep stage data extracted from 16 subjects were used for the training. The rest of sleep stage data from 7 subjects was used for testing. The recognition rate for the wake state could reach 98%. However, the recognition result for the first sleep stage was not satisfied. The recognition rate for the second sleep stage achieved 80.6%, and for the slow-wave sleep, the recognition rate reached 97.5%. It has shown the feasibility of automatic sleep classification using our proposed approach. We will increase the data samples to our experiments in the future work to improve the recognition rate.

誌謝 i
中文摘要 ii
Abstract iii
目錄 iv
圖目錄 vi
表目錄 viii
第一章 緒論 1
1.1前言 1
1.2研究動機與目的 1
1.3文獻回顧 2
1.3.1 國外文獻部分 2
1.3.2 國內文獻部分 4
第二章 研究背景 6
2.1 大腦生理結構 6
2.2 腦電波訊號 8
2.3 睡眠腦電波 10
2.4 Montage-組合範式 14
第三章 材料與方法 16
3.1 實驗對象與流程 16
3.2腦電波訊號量測 17
3.2.1 硬體架構 17
3.2.2 腦電波圖與肌電圖電極接法 17
3.2.3 腦電波訊號驗證 19
3.2.4 環境雜訊去除 20
3.3 類神經網路 21
第四章 實驗結果與討論 25
4.1 腦電波訊號切割 25
4.2 腦波特徵參數擷取 28
4.3 類神經網路訓練結果 32
4.4 辨識結果與討論 40
第五章 結論與未來展望 44
5.1 結論 44
5.2 未來展望 45
參考文獻 46
附錄 48

[1]Chiu, C.C, Yeh, S.J., Chen, C.H., ”Self-organizing arterial pressure pulse classification using neural networks: Theoretical considerations and clinical applicability” , Computers in Biology and Medicine, Vol. 30,2000, pp.71-88.
[2]許濬麟(2008)，”自我組織式類神經網路應用於即時心電圖波形分類”，逢甲大學自動控制工程研究所碩士論文.
[3]Backman, I.N., “Feature-based detection of k-complex wave in the human electroencephalogram using neural networks”, IEEE Transactions on Biomedical Engineering. Vol. 39, No. 12, December, 1992.pp.1305-1310.
[4]Shimada, T., “Sleep stage diagnosis system with neural network Analysis”, Proceedings of the 20th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Vol. 20, No 4, 1998.pp2074-2077.
[5]Park, H.J., Oh, J.S. et al., “Automated sleep stage scoring using hybrid rule and case-base reasoning”, Computers and Biomedicine Research, Vol. 33, 2000,pp. 330-349.
[6]Tian, J.Y., Liu, J.Q., ”Automated sleep staging by a hybrid system comprising neural network and fuzzy rule-base resonance”, IEEE Engineering in Medicine and Biology 27th Annual Conference ,2005.pp.4115-4118.
[7]Baumgart-Schmitt R, Herrmann WM et al., ”On the use of neural network techniques to analyze sleep EEG data”, Neuropsychobiology, Vol. 37,pp.49-58.
[8]Ebrahimi, F., ”Automatic sleep stage classification based on EEG signals by using neural networks and wavelet packet coefficients”, 30th Annual International IEEE EMBS Conference, 2005, pp.1151-1154.
[9]Shimada, T. , ”The automatic sleep stage diagnosis method by using SOM”, ICBME 2008, Proceedings 23, pp.245-248.
[10]Busek, P., ”Spectral analysis of heart rate variability in sleep”, Physiol. Res. 54, 2005, pp. 369-376.
[11]Lewicke, Arson, ”Sleep versus wake classification from heart rate variability using computational intelligence: Consideration of rejection in classification models“, IEEE transactions on biomedical engineering, Vol.55, NO.1, 2008.pp.108-118.
[12]溫國棟(2002)，”智慧化臨床資訊自動篩選與分析技術研究”, 中原大學醫學工程學系碩士論文.
[13]邱晧智(2008)，”以單頻道腦電波訊號偵測慢波睡眠“，國立中山大學機械與機電工程所碩士論文.
[14]Chang, W.H., “Basic mechanism for generation of brain rhythms”, Acta Neurological Taiwanica. Vol. 13, No. 4, 2004.pp.203-210.
[15]葉怡成，應用類神經網路，儒林，1997.
[16]Aguiar, P., EEG solver-brain activity and genetic algorithms, ACM Press, 2000. pp.88-84.
[17]Rechtschaffen, A., Kales, A., eds., ”A manual of standardized terminology, techniques and scoring system for sleep stages of human subjects ”. Los Angeles UCLA Brain Information Service/Brain Research Institute, 1968.
[18]Conn, P.P., Neuroscience in Medicine Third Edition, Humana Press, pp.626.
[19]關尚勇，林吉和，破解腦電波—EEG教材，藝軒圖書出版社,第二章，2007.
[20]林昇甫、洪成安, “神經網路入門與圖樣辨識”, 全華科技圖書股份有限公司.
[21Carpenter, G.A., Grossberg, S., “The ART of adaptive pattern recognition by a self-organizing neural network” , IEEE Computer, 1988. pp.77-88.
[22]Carpenter, G.A., Grossberg, S., “Pattern recognition by self-organizing neural networks”. The MIT Press, Massachusetts, Cambridge, 1991. pp. 397-424.