Recognizing Traffic Element Based on EEG Data at Gazing Point

The rapid and precise recognition of traffic elements during driving is essential for the reliable decision-making of autonomous driving systems. However, most existing models for traffic element recognition do not fully leverage the psychological and physiological characteristics of drivers, which limits the optimization potential of the system. To address this issue, this study introduces a traffic element recognition method based on eye-tracking and EEG data. In a simulated experiment, complex traffic driving scenario images with varied traffic elements from the Cityscapes dataset are used as stimuli, while EEG and eye-tracking data from the driver are simultaneously recorded. The gazing spatial points from eye-tracking data are utilized to label the traffic element categories from the images. Meanwhile, the EEG signal segments around the gazing temporal points are extracted as the input of the recognition model. Three machine learning models, including ANN, XGBoost, and LightGBM, are then employed to build the traffic elements recognition models based on EEG data at gazing point. The experimental results demonstrate that the proposed framework significantly improves the accuracy of traffic element recognition, with XGBoost model achieving the best recognition accuracy of 91.8%. By integrating the cognitive mechanisms of human-beings when facing complex traffic scenarios, this work provides a crucial theoretical foundation and practical support for the rapid recognition of traffic elements in intelligent driving systems.