A Review on Deep Learning Techniques for EEG-Based Driver Drowsiness Detection Systems

Driver Drowsiness is considered one of the significant causes of road accidents and fatal injuries. Due to this, creating systems that can monitor drivers and detect early drowsiness has become an important field of research and a challenging task in recent years. Several research attempts were proposed to solve this problem based on several approaches and techniques. The Electroencephalogram (EEG) is one of the most efficient and reliable method, among the physiological signals-based monitoring approaches. In this area, many Machine Learning (ML) techniques have been used to detect EEG-based driver drowsiness. However, due to the limitations of ML techniques, many researchers have shifted their focus to the use of deep learning (DL) techniques, which have demonstrated superior performance in many fields including the physiological signals classification tasks. This paper reviews and discusses numerous new research papers that have proposed and implemented driver drowsiness detection systems based on EEG and deep learning techniques. In addition, we have outlined the limitations and difficulties of the existing works and highlighted and proposed some propositions that will help future field researchers enhance and generalize the results. Based on our thorough analysis, we have determined that the latest advancements in detecting driver drowsiness have employed the convolutional neural network (CNN) technique, which has demonstrated effective performance in classifying signals. Furthermore, the primary issue encountered in all works is developing a more precise and accurate method. Nevertheless, we seek a precise system capable of swiftly identifying a state of drowsiness while using minimal spatial memory and processing resources.