Driving Behavior Recognition Based on EEG Channel Attention Mechanism
Electroencephalogram(EGG)signals,with their high temporal resolution among other advantages,have become an essential tool for recognizing drivers'cognitive states and assessing driving performance.Previous research on brain electrical activity in the context of driving behavior has often been limited to abnormal driving states,such as fatigue detection and distracted driving,neglecting normal driving scenarios.This paper focuses on regular driving behaviors recognition.Through driving simulation experiments,this study synchronously collected driving and brain electrical activity data from drivers while they performed acceleration,deceleration,and turning maneuvers.A channel attention-separable convolutional neural network based on the squeeze-and-excitation module is constructed to recognize the driving behaviors and optimize channel selection across individuals'brain electrical signals.The results show that the proposed model achieved an accuracy of 82%in recognizing three types of driving behaviors while reducing the number of channels by 70%without compromising prediction accuracy.The effectiveness of the model was demonstrated through ablation experiments and comparisons with other baseline models.Analysis of the optimal channel combinations'scalp topography revealed that the frontal and occipital areas of the brain are most relevant to regular driving behaviors.The findings of this study provide a methodological basis to understand driving behavior from a cognitive perspective and for brain-like driving decision-making.
NETL
NSTL
万方数据
