發展EMD方法濾除腦波眨眼訊號並應用於測量疲勞狀態之研究

Abstract

腦波是檢測大腦活動的重要工具，但在利用腦波做各項研究和應用時，受試者的眨眼訊號常常會成為干擾研究的大麻煩，過去已有不少研究者發展濾除眼動雜訊的數學方法，但各有其限制和缺點，其中最廣為應用的是Independent Component Analysis (ICA)。本論文試圖發展可用性更高的眨眼訊號濾除法，利用Empirical Mode Decomposition (EMD)將訊號分解，只取前五個Intrinsic Mode Function (IMF)分析，丟棄其他的IMF和最後的剩餘訊號，以去除腦波中的低頻訊號；利用適當組合並移除前五個IMF中振幅過大的振動以去除腦波中的高振幅訊號，最後再將處理過的訊號相加，運用這兩種訊號移除機制可大致上去除眨眼雜訊而保留腦波訊號，若其他干擾訊號具有高振幅或低頻率的特性，則此方法可將之一併濾除。此眨眼訊號濾除法和過去的方法相較之下，限制較少，可應用於單通道腦波機；濾除機制較直覺，可明確了解何種訊號在此過程中會被濾除；計算速度快，節省計算資源；方便性高，不需針對受試者調整參數；方法新穎，與前人常用的眨眼或眼動訊號濾除機制有很大的差異性，這些優點使得本方法在眨眼訊號濾除工作具有更好的可用性和方便性。本論文並運用此眨眼訊號濾除方法以腦波做疲勞度偵測之研究，此方法在實驗中可有效的濾除眨眼訊號和大部份的低頻干擾訊號，亦減輕其他肌電訊號對實驗造成的影響，藉由此眨眼訊號濾除法令本研究可分析受試者張眼狀態下之腦波，不致因干擾訊號的介入使腦波訊號無法分析。研究方法是讓受試者做事先設計好的電腦作業，在作業前、後量測受試者的當下狀態腦波並用問卷測得受試者的主觀疲勞度，若假設受試者電腦作業前、後的狀態變化主要來自疲勞度的改變，則比較受試者在電腦作業前、後的腦波變化和主觀疲勞度變化，可適當的了解疲勞度和腦波的變化關係，實驗收集了25位受試者的腦波資料，以過零點法、修正後的過零點法、傅利葉轉換三種方式定義的腦波平均頻率與腦波功率(power)和問卷測得的主觀疲勞度做統計比較，可發現數種向度的主觀疲勞度和不同腦區的不同腦波特徵具統計相關的有效性，「我沒興趣做以前感興趣的事」此一主觀疲勞向度甚至與右腦提取之腦波特徵具極高相關性。

Electroencephalogram (EEG) is important in examine brain activities. Among the many sources of artifacts in EEG recording, eye activity plays a dominate role. There are many methods developed to remove it, but most of them have their shortcomings or restrictions. The most frequently used method to remove ocular artifacts is independent component analysis (ICA), but it can be applied in multi channel EEG data only. A new method was presented in this thesis to remove eye blink artifacts from EEG data which is based on the mathematical method called empirical mode decomposition (EMD). The method has several advantages that make it more convenient and flexible compared with other methods. It could be used in single channel EEG data and needn’t modify personal parameters to fit each data. When applying the method, use EMD to decompose EEG data into sever intrinsic mode functions (IMF) and one residue first, and then remove artifacts liked data by retaining IMF from first to fifth only and eliminating over limit vibration of these IMF. The thesis also utilizes this EMD based eye blink signal removing method in an experiment which designed to estimate fatigue with EEG. This method works effectively in the experiment in removing artifacts, not only eye blink artifacts but also other kinds of artifacts. The design of this experiment needs 25 subjects doing a planed work, and then measuring subjective fatigue transfer and EEG pattern transfer between non-worked and worked states. A fatigue scale was used in estimating fatigue levels. A brain wave machine was used to measure EEG pattern. After getting the data, this research tries to find connections between EEG and fatigue with correlation of statistics.