An EEG-based machine learning system classified driver mental state in real-time with 96.5% accuracy and predicted driving performance before driving with 85% accuracy.
An EEG-based machine learning system can accurately classify driver mental states and predict driving performance, potentially enhancing automotive safety.
Human error is considered one of the major causes of car accidents. One potential approach to reduce human driving errors is to continuously monitor the driver's performance while driving. This could help in detecting potential risks and thus reduce the likelihood of accidents. In this paper, we introduce a machine learning system that analyzes the driver's brain activity to monitor and predict the driver's performance. While driving, the system monitors the driver's mental state by analyzing acquired Electroencephalography (EEG) signals. Additionally, the proposed system acquires EEG activity from the driver before driving and predicts the driving performance along the intended route. The proposed system is tailored for the Automotive Open System Architecture (AUTOSAR) framework. Our results demonstrate the ability of the system to classify the mental state of the driver in real-time into three states (focused, unfocused, and drowsy) with a mean accuracy of 96.5% across three examined subjects. The system also predicts the driver's performance before driving from the recorded EEG signals with a mean accuracy of 85%. These results indicate the utility of EEG signals analysis in enhancing the safety of futuristic automotive applications.
Elsherif et al. (Tue,) conducted a other in Driving performance (n=3). EEG-based machine learning system was evaluated on Classification of driver mental state (focused, unfocused, and drowsy). An EEG-based machine learning system classified driver mental state in real-time with 96.5% accuracy and predicted driving performance before driving with 85% accuracy.