基於多重生理訊號參數之即時無線瞌睡偵測系統

Abstract

近年來，交通意外是一個造成駕駛死亡的至關重要原因，其中駕駛者的精神狀況不佳所造成車禍意外佔了絕大多數比例，所以開車駕駛瞌睡監控問題是我們嘗試克服之處，試著以人為方式來減少車禍發生。近年來相關的開車監控研究引進了生理參數來做為開車即時瞌睡狀況的比較依據，如心電圖、眼電波圖、肌電圖或腦波圖等，較影像辨識來得直接與精確，使用者可以不必受影像定位之問題影響，本論文即針對生理參數中之腦波以及眼電波參數做進一步的探討。我們設計了一套無線可攜式的多重生理訊號擷取系統以及包含生理回饋機制如電刺激器等的數位訊號處理平台，再搭配非監督式分析演算法來做即時的瞌睡判斷。使用非監督式演算法的優勢在於可移除掉不同人、不同次測量中個別跟環境差異性。本論文藉由虛擬實境模擬環境所記錄下開車偏移量來當作瞌睡程度的參考，並與所發展的非監督式分析法的相互比對關係來證明此演算法對瞌睡程度偵測的功效與可行性，最後實現在數位訊號處理平台上。經由實際測試，可以成功在駕駛者有睡意時，利用電刺激器或是警示音提醒駕駛保持清醒，確保開車時的安全。

In recent years, traffic accident is one of the critical reasons to cause deaths of drivers. Drivers’ drowsiness has been implicated as a causal factor in many accidents because of the marked decline in drivers’ perception of risk and recognition of danger, and diminished vehicle handling abilities. Consequently, if the mental state of drivers could be real-time monitored, drowsiness detection and warning could effectively avoid disasters such as vehicle crashes in working environments. Some previous researches used non-physiological method, as eye closure with CCD image tracking, such as the pupil recognition, blink detection or identification of the drivers head shaking frequency. However, for CCD image tracking, users couldn’t move for free, and the images detecting performance were easily be interfered by external flash light. Other studies used physiological parameters to increase the accuracy of drowsy detection, like pulse wave analysis with neural network, electrooculography (EOG), electromyography (EMG), and electroencephalogram (EEG) measurement. In this study, we proposed a real-time wireless system for drowsiness detection. A wearable, wireless and real-time bio-signal acquisition system was designed for long-term monitoring. In the other hand, not only EEG but also EOG signals were acquired by our system to increase the accuracy of drowsiness detection. Furthermore, an algorithm of drowsiness detection was also proposed to reduce the computation complexity, and was implemented in a portable DSP module with bio-feedback as bio-stimulator or buzzer. In order to estimate the level of drowsiness, a lane-keeping driving experiment was designed and the drowsiness level of drivers was indirectly assessed by the reaction time under Virtual Reality Driving Simulation Environment. The advantage of this unsupervised algorithm can remove the differences between individual and environment in different people or measurements. For the purpose of verifying the accuracy and feasibility of our proposed unsupervised algorithm, drowsiness status estimated by driving performance was compared with the results obtained by our proposed unsupervised algorithm. The results of comparison showed that our algorithm can detect driver’s drowsiness status precisely. In addition, our system can be successfully applied in practice to prevent traffic accidents caused by drowsy driving.