People frequently struggle to juggle their work, family, and social life in today's fast-paced environment, which can leave them exhausted and worn out. The development of technologies for detecting fatigue while driving is an important field of research since driving when fatigued poses concerns to road safety. In order to throw light on the most recent advancements in this field of research, this paper provides an extensive review of fatigue driving detection approaches based on electroencephalography (EEG) data. The process of fatigue driving detection based on EEG signals encompasses signal acquisition, preprocessing, feature extraction, and classification. Each step plays a crucial role in accurately identifying driver fatigue. In this review, we delve into the signal acquisition techniques, including the use of portable EEG devices worn on the scalp that capture brain signals in real-time. Preprocessing techniques, such as artifact removal, filtering, and segmentation, are explored to ensure that the extracted EEG signals are of high quality and suitable for subsequent analysis. A crucial stage in the fatigue driving detection process is feature extraction, which entails taking pertinent data out of the EEG signals and using it to distinguish between tired and non-fatigued states. We give a thorough rundown of several feature extraction techniques, such as topology features, frequency-domain analysis, and time-domain analysis. Techniques for frequency-domain analysis, such wavelet transform and power spectral density, allow the identification of particular frequency bands linked to weariness. Temporal patterns in the EEG signals are captured by time-domain features such autoregressive modeling and statistical moments. Furthermore, topological characteristics like brain area connection and synchronization provide light on how the brain's functional network alters with weariness. Furthermore, the review includes an analysis of different classifiers used in fatigue driving detection, such as support vector machine (SVM), artificial neural network (ANN), and Bayesian classifier. We discuss the advantages and limitations of each classifier, along with their applications in EEG-based fatigue driving detection. Evaluation metrics and performance assessment are crucial aspects of any detection system. We discuss the commonly used evaluation criteria, including accuracy, sensitivity, specificity, and receiver operating characteristic (ROC) curves. Comparative analyses of existing models are conducted, highlighting their strengths and weaknesses. Additionally, we emphasize the need for a standardized data marking protocol and an increased number of test subjects to enhance the robustness and generalizability of fatigue driving detection models. The review also discusses the challenges and potential solutions in EEG-based fatigue driving detection. These challenges include variability in EEG signals across individuals, environmental factors, and the influence of different driving scenarios. To address these challenges, we propose solutions such as personalized models, multi-modal data fusion, and real-time implementation strategies. In conclusion, this comprehensive review provides an extensive overview of the current state of fatigue driving detection based on EEG signals. It covers various aspects, including signal acquisition, preprocessing, feature extraction, classification, performance evaluation, and challenges. The review aims to serve as a valuable resource for researchers, engineers, and practitioners in the field of driving safety, facilitating further advancements in fatigue detection technologies and ultimately enhancing road safety.
