| 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。深入探討造成事故中的原因，發現肇事原因所佔的
30%比例與駕駛人員疲勞有關。本研究主要目的，在於汽車駕駛人員疲勞檢測系統的研發，藉由架設汽車駕駛模擬環境，提出以腦波( EEG
)為主軸之疲勞瞌睡研究方法，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，在不同受測者疲勞瞌睡狀態有其最佳之閥值識別參數。
關鍵字 : 疲勞駕駛、腦波、駕駛參數、經驗模態分解法、時頻分析、支持向量機
Every year, the number of
deaths due to traffic accidents reaches 600 thousand, causing
serious economic losses indirectly. In depth study of the causes
reasons of the accidents, it was found that more than thirty
percents (30%) is related to leading to drivers mental-fatigue.
The main purpose of this thesis is developing car drivers
mental-fatigue detection system research and development by
setting up car driving simulation environment. Proposed to use
Electroencephalogram (EEG) as the main research methods of
mental-fatigue and research the reaction of car driver to the
driving parameters of the dynamic behavior, to look for the key
feature between driving parameters and EEG before driver
mental-fatigue state. In the EEG electrode measurement
experiment used in this thesis two-channel EEG-based system,
existing eye-blinking artifacts interference problem. In order
to solve eye-blinking artifacts problem, This thesis is based on
empirical mode decomposition (EMD) to remove eye-blinking
artifacts from EEG signals, and restoring the original EEG
signals measured. When the subjects were driving steering-pedal
simulator for simulation road environment will be measured after
the time-frequency analysis of EEG by comparison, can be found
in the mental-fatigue state during high frequency (8~12 Hz) band
are part of the power burst phenomenon. Driving parameters can
be found some subjects in the EEG corresponds to the
mental-fatigue state, it happens around the steering wheel
shaking and swing phenomenon. According to time-frequency
analysis of EEG, this thesis will be high frequency (8~12 Hz)
band power burst as mental-fatigue of feature extraction and
support vector machine (SVM) classifier for threshold feature
identification and classification. Identified by the SVM
classifier found that high V frequency (8~12 Hz) band
power really be able to view the subjects of mental-fatigue
state, and in different subjects have the best recognition
threshold parameters.
Key words: Fatigue Driving,
EEG, Driving Parameters, EMD, Time-Frequency Analysis,
SVM
|
