Design of a driver fatigue detection and graded warning system based on multi-feature fusion

Fatigue driving can cause a decrease in the driver's perception, judgment, and decision-making abilities, which is a major safety hazard in traffic safety. In order to effectively detect the driver's fatigue status and achieve graded warning, and reduce the traffic accident rate, this study takes the behavioral characteristics of drivers in a fatigued state and the graded early warning mechanism as the research object, develop a multimodal fatigue detection system using infrared vision sensors and integrated multi-source data fusion to holistically monitor driver fatigue. A spatio-temporal based convolutional neural network is adopted to identify the driver's fatigue state, and the features of the driver's PERCLOS value, yawning frequency, and head posture are fused and trained to provide a comprehensive judgment of fatigue. The experimental results verify that this detection method can accurately determine whether the driver is experiencing fatigue driving, and issue warning measures such as sound and vibration based on the degree of fatigue, effectively reducing the probability of fatigue driving and ensuring the safety of drivers and passengers to a certain extent.