Drowsy driving is a major contributor to serious traffic accidents, highlighting the urgent need for effective real-time detection systems. This study proposes a real-time drowsiness detection system based on electroencephalogram (EEG) signals and a lightweight convolutional neural network (CNN). The system comprises five main components: EEG signal acquisition, preprocessing, feature extraction, CNN-based classification, and user feedback delivery via an Android application. The experiment involved four healthy adult male participants with an average age of 24.5 years. EEG data were collected using the DSI-24 device, and the relative power in the alpha band from the prefrontal (Fp1, Fp2) and occipital (O1, O2) regions was identified as the primary feature for distinguishing drowsiness. The proposed CNN model, trained on these features, achieved a classification accuracy of 91.56%, comparable to the 92.66% accuracy of the more complex AlexNet model, while being significantly more lightweight and suitable for real-time deployment on embedded systems. The Android application provides real-time feedback on the user’s drowsiness level and recommends nearby rest areas to help mitigate the risk of drowsy driving. This study presents a practical and efficient EEG-based driver monitoring solution. Future work will focus on large-scale data collection under actual driving conditions to further validate and improve the system’s performance.
Ssang-Hee Seo (Tue,) studied this question.