- 學位論文
基於眼動圖與肌電圖之自動睡眠分期系統
Automatic Sleep Stage Recognition System Based on EOG and EMG
摘要
近年來由於時代的變遷及科技的進步，造成現代人生活忙碌，工作時間倍增、壓力變大，很多人會犧牲睡眠時間完成手邊的工作，睡眠品質下降，不只會導致專注力下降、影響情緒、工作效率不佳等等的結果，甚至有可能引貣交通意外。 本論文提出基於眼動圖與肌電圖之自動睡眠分期系統，希望在進行睡眠檢查時，減少所需量測項目，僅使用眼動圖和肌電圖的生理訊號進行睡眠判讀，達到降低成本、減少電極片與訊號線的目的。實驗資料取自於Sleep-EDF(Sleep Recordings and Hypnograms in European Data Format)資料庫，8位受測者之眼動圖與肌電圖經過訊號前置處理與特徵擷取之後，建立一決策樹，以二元分類法之方式，透過三個分類器根據各個睡眠期之生理特徵分別設定不同門檻值，依序將快速動眼期、慢波睡眠期、清醒期與淺眠期辨識出來。 在實驗結果中，單獨以分類器效能而言，快速動眼期分類器的kappa值為三者中最佳，慢波睡眠期為次佳，再來才是清醒期，kappa值分別為82.46%，77.77%，75.35%，平均為78.5%，但若透將分類器進行不同排序，將三分類器中kappa值最 高之快速動眼期分類器進行第一階段篩選，篩選後的資料再經由次佳與次次佳之慢波睡眠器與清醒期分類分類器進行分析，可將kappa值提升到81.7%。
並列摘要
Sleep is very important for a person’s physical and mental health. We spend one third of our life. Most people don’t attach much importance to sleep, even ignored sleep problem. In this paper, we describe attempts at constructing an automatic sleep stage recognition system based on EOG (Electrooculogram) and EMG (Electromyography). Wake stage (WS), light sleep (LS), slow wave sleep (SWS) and rapid eye movement (REM) were classified by decision tree with three classifiers. The EOG horizontal and EMG submental data used for this analysis was obtained from the Sleep-EDF (Sleep Recordings and Hypnograms in European Data Format) database. The concordance between Sleep-EDF database and recognition was evaluated by Cohen’s Kappa index. In results, the kappa coefficient of REM classifier, SWS classifier and WS classifier is 82.46%, 77.77% and 75.35%. Besides, we arranged three classifiers in different orders. We raised up the average of kappa coefficient to 81.7% from 78.53% when we arranged REM classifier to be a first order classifer, SWS classifier and WS classifier is second and third order.