Driver Distraction State Recognition Based on Minimum Spanning Tree Features using EEG Data

Driving distraction is recognized as a significant factor in traffic accidents. In recent years, neurobiological methods have emerged as a promising approach for intelligent distraction detection. This study proposes a novel method, leveraging collected EEG data related to driving distraction, to capture diverse minimum spanning tree topologies, thereby supporting the identification of driving distraction states. The minimum spanning tree is derived from the functional connectivity matrix, with quantifiable node attributes along with global attributes computed. These features serve as inputs for training support vector machines with radial basis functions and random forests to classify distracted and non-distracted driving states. Results indicate that different feature representations provide complementary information, achieving an 85.42% accurate recognition of distracted driving states in the support vector machine. Indicating that the adoption of the minimum spanning tree as an indicator for driving distraction is effective, this approach can provide a technical foundation for intelligent driver distraction warning systems.