مهسا محمدشريفي داراني

Drowsiness is defined as a transitional state from wakefulness to sleep, which can significantly impair human al‎e‎rtness. It can notably increase the risk of accidents in driving and other activities. Therefore, the development of precise and reliable methods for detecting this condition is of paramount importance. Recent advancements in deep learning and the utilization of physiological signals such as electroencephalogram (EEG) signals have facilitated more accurate detection of drowsiness. The use of deep learning methods enables automatic extraction of complex features from EEG signals, enhancing the capability and efficiency of drowsiness detection. Hence, this research proposes an innovative method for drowsiness detection based on EEG signals using deep learning-based networks. By examining the limitations and weaknesses of conventional signal processing methods and classical classification algorithms, this study presents novel approaches. The proposed method achieves higher accuracy and detection capability by eliminating signal processing methods, employing various neural networks, and designing a new architecture for generating more data to balance the dataset. In this thesis, three models with different structures for drowsiness detection and sleep stages are presented, utilizing EEG signals directly as input. The first and second proposed models are designed for drowsiness detection, while the third proposed model focuses on sleep stage detection to improve the N1 stage, associated with drowsiness. The performance of these models is eva‎luated on different EEG signal channels from two sleep datasets. eva‎luation results demonstrate that the proposed models exhibit high accuracy and efficiency, with the best proposed model achieving an overall accuracy of 95.62%, providing the best performance in drowsiness detection.

خواب آلودگي , مراحل خواب , سيگنال فيزيولوژيكي , شبكه عصبي , يادگيري عميق

Drowsiness , Sleep Stages , Physiological Signal , Neural Network , Deep Learning

Mahsa Mohammad Sharifi

Dr. Ahmad Ayatollahi