Using single-channel EEG for detecting drowsiness by deep learning techniques

Living standards are rising, leading to increased road traffic and a subsequent increase in traffic accidents, especially those caused by driver fatigue or drowsy driving. It is estimated that around ۷ percent of accidents and ۱۶.۵ percent of fatal crashes result from drivers falling asleep, emphasizing the need for a drowsiness detection system. One approach for detecting drowsiness involves analyzing Electroencephalography (EEG) signals, and there is limited research using convolutional neural network (CNN) architectures and deep learning algorithms in this area. This study deals with a publicly available dataset providing EEG data from ۱۱ individuals. In this study, ۵ channels were compared and only one EEG channel was used to detect drowsiness. Previous research utilized time resolution above ۳ seconds, but in this work one-second resolution was used, allowing for rapid recognition of drowsy states in drivers or other high-risk tasks without calibration. The study's results revealed that the Fz channel achieved the best outcome with an accuracy of ۷۰.۳۵ percent in one-second time segments.

کلیدواژه ها:

Single channel EEG ، brain drowsiness ، convolutional neural network ، leave-one-out cross-validation ، deep learning