On identification of driving-induced stress using electroencephalogram signals: A framework based on wearable safety-critical scheme and machine learning

Driving an automobile under high stress level reduces driver's control on vehicle and risk-assessment capabilities, often resulting in road accidents. Driver's anxiety therefore is a key factor to consider in accident prevention and road safety. This emphasizes the modern computing techniques to assist drivers by continuous stress level monitoring. Development of such a system requires designing a framework, which can recognize the drivers' affective state and take preventive measures to account for escalating stress level. This work presents a machine learning-based approach to identify driving-induced stress patterns. For this, electroencephalograph (EEG) signals are utilized as the physiological signals. The ongoing brain activity is logged as EEG signal to determine the link between brain dynamics and emotional states. Three classifiers are utilized in this work, namely: Support Vector Machine (SVM), Neural Network (NN), and Random Forest (RF) to classify EEG patterns on the basis of the subject's self-reported emotional states while driving in various situations. A framework is proposed to recognize emotions based on EEG patterns by systematically identifying emotion-specific features from the raw EEG signal and investigating the classifiers' effectiveness. A comprehensive analysis of various performance measures concludes that among the three classifiers employed in this study, SVM performs better to distinguish between rest and stress state. The evaluation obtained an average classification accuracy of 97.95% +/- 2.65%, precision of 89.23%, sensitivity of 88.83%, and specificity of 94.92%; when tested over 50 automotive drivers.