Decoding Braking Intentions From EEG Signals: A Convolution-Transformer Network Approach

By monitoring the driver’s electroencephalogram (EEG) signals while braking, it is feasible to acquire braking intentions more rapidly compared to detecting braking maneuvers. Having an advantage in time is crucial when there is a high possibility of a traffic collision occurring. The research involves conducting studies on three distinct types of EEG signals from drivers: emergency braking, soft braking, and normal driving. For decoding intentions, we propose a network structure for decoding braking intentions from drivers’ EEG signals. It integrates the advantages of convolution and Transformer. The network combines local and global features in a unified EEG classification framework. We conducted experiments on both our own collected dataset and the publicly available dataset and achieved up to 90% or more accuracy in recognizing emergency braking intent and up to about 80% accuracy in recognizing soft braking intent. This work can additionally enhance the advancement of EEG-based human-centric supported driving systems, aiming to enhance driving safety and comfort.