Graph-based Driving Style Recognition from Electrophysiological Analysis during Car Following

Driving styles can be used to characterize the behavioral characteristics of drivers, and provide a personalized foundation for the development of intelligent driving systems. The electrophysiological analysis is one of the emerging technologies for driving style recognition. However, most studies focused on classifying driving styles based on electroencephalogram signals have predominantly extracted the time-frequency features, which ignore the spatially coordinated activities of multiple brain regions. Therefore, this paper proposes a graph-based driving style classification method by implementing electrophysiological analysis on brain functional connectivity networks. By adopting graph convolution network on the time-frequency and functional connectivity features, the constructed model attains an accuracy of 88.3% in recognizing three car-following driving styles. This approach can help to promote the electrophysiological recognition of driving styles with a network perspective.