Intelligent Fatigue Driving Detection Method Based on Fusion of Smartphone and Smartwatch Data

Fatigue driving is one of the primary causes of traffic accidents, and effective detection of driver fatigue can help reduce the occurrence of such accidents. This study conducted real-world vehicle driving experiments, where physiological data and vehicle trajectory data of 15 drivers were collected using smart wearable watches and smartphone applications. Data features were extracted, and an Attention-LSTM deep neural network was employed to establish a normative driving model for each participant, enabling intelligent detection of fatigue driving states. Two approaches were employed for model construction, one utilizing only trajectory data features and the other fusing trajectory and physiological data features. The results indicated that the model based on the fusion of trajectory and physiological features outperformed, with an average detection F1 score of 80.31%, average precision of 79.27%, and average recall of 82.71%. The findings of this study suggest promising applications for fatigue driving detection in real-world driving scenarios.