Replacing EEG with peripheral physiological signals for driving behavior recognition: a physiology-informed multimodal approach

Electroencephalography (EEG) has been widely used for driving behavior recognition due to its ability to capture neural activity; however, EEG acquisition typically requires expensive hardware and complex instrumentation, which limits its practicality in real-world applications. This study investigates whether low-cost peripheral physiological signals can serve as an effective alternative to EEG for driving behavior recognition. A publicly available multimodal driving dataset collected in a simulator environment was analyzed, containing physiological recordings from 35 participants performing five driving behaviors, including smooth driving, acceleration, deceleration, lane change, and turning. Three peripheral physiological signals-electrocardiography (ECG), electromyography (EMG), and galvanic skin response (GSR)-were integrated using a physiology-informed multimodal learning framework designed to capture complementary physiological responses and temporal dynamics across modalities. Experimental results show that the proposed peripheral multimodal configuration (ECG+EMG+GSR) achieves an average accuracy of 0.795 and macro-F1 score of 0.805 under five-fold cross-validation, with MCC and Cohen's kappa reaching 0.731 and 0.730, respectively. Controlled comparisons with six representative baseline methods further showed that the proposed framework consistently outperformed all competing methods across different physiological modality configurations. Notably, the ECG+EMG+GSR configuration outperformed the EEG-only setting while avoiding EEG acquisition and its associated multi-electrode setup, thereby providing strong recognition performance with substantially lower sensing burden and deployment complexity. These findings highlight the potential of multimodal peripheral physiological sensing as a practical and cost-efficient alternative to EEG for driving behavior recognition in intelligent transportation systems.