動態權重調整之整合系統應用於駕駛者反應時間預測

Abstract

車禍是造成美國人口死亡的主要原因之一。大部分的車禍是由於人們操作和行為上的失誤所造成的，尤其是當開車者疲勞或是注意力不集中的時候。很明顯的，去預測駕駛者的認知行為亦或更進一步的去建造一個駕駛對於危險的反應時間預測模型是一项急迫的研究。在過去二十年間，由大腦所產生的電場的EEG(electroencephalogram)訊號被認為是了解人類認知行為很穩固的生理資訊指標。然而，個體之間的差異或是受測者本身的內部差異例如疲勞、壓力或是注意力不集中都可能造成人們大腦基準狀態的改變進而影響開車行為表現以及大腦活動之間的關係，造成難以使用一般化的預測模行去對反應時間去做預測。為了解決這個困難，我們提出了一個整合學習之動態權重調整預測系統的架構。這個系統是由眾多子預測模型所組成的，其中，每一個子預測模型是由相似的EEG-RT對應關係的群集所訓練而成的，每一個Model會預測出受測者在下一個時間點的反應時間。每一個子預測模型的權重是由EEG Theta頻帶和coherence所訓練出的GMM模型所給出的事後機率所決定的，由結果顯示出在預測開車行為表現上，我們所提出的時變動態權重調整系統明顯優於傳統的預測系統。因此，我們所提出的適應性調整系統也經由實驗證實是可以解決個體差異化所導因的大腦和開車行為的變化的不同步。這個研究的主要貢獻在於使用整合式EEG-based適應性方法來增加危險警示系統的準確度。

Motor vehicle crashes are the leading cause of fatalities in the US. Most of these accidents are caused by human mistakes and behavioral lapses, especially when the driver is drowsy, fatigued, or inattentive. Clearly, predicting a driver’s cognitive state, or more specifically, modeling a driver’s reaction time (RT) in response to the appearance of a potential hazard warrants urgent research. Recently, the electric field that is generated by the activity of the brain, monitored by an electroencephalogram (EEG), has been proved to be a robust physiological indicator of human behavior. However, mapping the human brain can be extremely challenging, especially owing to the variability in human beings over time, both within and among individuals. Factors such as fatigue, inattention and stress which can induce homeostatic changes in the brain, which affect the observed relationship between brain dynamics and behavioral performance, and thus make the existing systems for predicting RT difficult to generalize. To solve this problem, an ensemble-based weighted prediction system is presented herein. This system comprises a set of prediction sub-models that are individually trained using groups of data with similar EEG-RT relationships. The prediction outcomes of the sub-models are predicted by the weights that are derived from the EEG theta power and coherence, whose changes were found to be indicators of variations in the EEG-RT relationship, to obtain a final prediction. The results thus obtained reveal that the proposed system with a time-varying adaptive weighting mechanism significantly outperforms conventional systems in modeling a driver’s RT. The adaptive design of the proposed system demonstrated its feasibility in coping with the variability in the brain-behavior relationship. The contribution of this work is that simple EEG-based adaptive methods are used in combination with an ensemble scheme to increase significantly system performance.