A Transfer Learning-Based Approach for Fatigue Detection Through the Fusion of Physiological Signals

The improvement of road safety and the reduction of accident occurrences globally necessitates the detection of driver fatigue. This article explores the current research gap in identifying the most effective physiological signals for fatigue detection. Through a comprehensive analysis, physiological signals are converted into images and processed using pretrained deep neural networks (DNNs) to detect patterns indicative of fatigue. This research stands as a pioneering effort in fatigue detection, examining various combinations of physiological signals, pretrained DNNs, and classification methods within a dataset. Our thorough evaluation demonstrates that multiple classifiers using ECG signals can achieve near-perfect performance metrics (accuracy, area under the curve, precision, and recall) on test sets. These highlight the transformative potential of artificial intelligence (AI) technologies in redefining current systems for monitoring driver fatigue, marking a significant leap forward in the application of AI for ensuring driver safety and preventing accidents caused by fatigue. The findings reinforce the critical role of incorporating advanced AI techniques into driver-assistance systems to develop more reliable and efficient fatigue detection solutions.