摘要
全世界每年因交通事故導致的死亡人數達60萬人，間接造成嚴重的經濟損失。在許多事故調查報告中指出，其肇事原因所佔的 30% 比例與駕駛人員疲勞有關。本論文主要目的，在於汽車駕駛人員疲勞檢測系統的研發。此系統包含道路駕駛模擬環境建置與採用分析腦波的方法作為駕駛疲勞的識別，並分析能具體反應駕駛人動態行為的駕駛參數，尋找疲勞瞌睡前駕駛參數與腦波瞌睡間的關鍵特徵。腦波量測應用於在本研究實驗之雙通道腦波機制，普遍存在著眨眼訊號干擾問題。為了解決眨眼訊號之干擾，本研究利用經驗模態分解法(EMD)，將腦波中存在的眨眼訊號給予濾除，解離出所需量測之原始腦波訊號。當受測者進行道路模擬駕駛工作時，將量測後的腦波訊號經由時頻分析比較，可以發現在疲勞瞌睡期間，腦波頻率為高頻 8~12 Hz 波段的部份有功率脈衝現象。另在駕駛參數方面，可發現部分受測者在對應腦波之疲勞瞌睡狀態時，其方向盤有明顯左右偏擺現象。根據腦波時頻分析結果，本研究將以腦波高頻 8~12 Hz 波段的功率脈衝作為疲勞瞌睡之特徵擷取，並利用支持向量機(SVM)分類器進行閥值特徵的識別與分類。經由分類器識別結果發現，其腦波高頻 8~12 Hz 能量功率確實可以檢視受測者之精神狀態。本研究結果對於個體精神狀態的識別可供未來相關研究進一步的發展與應用。
並列摘要
The number of deaths due to traffic accidents reaches 600 thousand annually. That indirectly causes a serious economic loss. Many studies indicate that the accident due to the driver fatigue are more than 30% leading to fatality. An objective of this thesis is to develop a system in detecting driver fatigue. The research includes a driving virtual environment and adoption of method to analyze Electroencephalogram (EEG) signals. The driver’s behavior is revealed by detected driving parameters and the brain signal from conscious to microsleep. The brain signal includes the unexpected signal from eye-blinking which will be decomposed by the method of empirical mode decomposition (EMD). After EMD processing to eliminate unwanted noise, the signals between 8 to 12 Hz are found at obvious power peaks during the micro-sleep, meanwhile the recorded steering wheel parameter is also found an distinctly sharp change. These peaks can be discriminated by support vector machine (SVM). This study is paving a way to spot the area of mental state identification.