Investigation and Analysis about Multimodal Fatigue Detection

Fatigue driving accounts for a significant proportion of traffic accidents. Multimodal fatigue detection offers a new effective way to detect fatigue. This article selected six articles by some criteria and reviewed them carefully. The improvements of the method include using lightweight YOLOv8 networks to process complex images; others process multimodal EEG signals via non-smooth non-negative matrix factorization (nsNMF) and Gramian angular field imaging. Attention-based networks like MMA-Net and TMU-Net are designed to fuse EEG, EDA, PPG, and EOG signals. Additionally, LSTM-based models analyze PPG and facial features, while enhanced MTCNN and PFLD algorithms improve detection accuracy and reduce individual variability. Based on the methods, the article summarizes some challenges in the multimodal fatigue detection and proposes the future of the multimodal fatigue detection. In general, the article provides a comprehensive review to guide further research about multimodal fatigue detection.