Eye States Detection from Wearable EEG Sensing for Driver Safety

With the increasing interest in wearable electroencephalogram (EEG) devices, there is a growing interest in developing these technologies for real-time monitoring of driver safety. This study focuses on the development of a novel approach for detecting eye states, such as drowsiness, using EEG signals obtained from a wearable device. By analyzing the electrical activity of the brain, particularly in regions associated with visual attention and alertness, our proposed system aims to provide a reliable and non-invasive method for assessing a driver’s cognitive state. For that, Oz electrode is selected for eye close (EC) and eye open (EO). We integrating theta, alpha, and beta bands extracted from one second detrended EEG signal. In addition, we calculate temporal features using sample entropy (SampEn) and spectral features using Power Spectral Density (PSD). Finally, Convolutional neural network (CNN) is trained from the selected features to provide the estimates of EC and EO by utilizing their ability to detect hidden features in complex EEG data. The study also provides the insight into the importance of features from the frequency bands for EO and EC detection.