How EEG-based cross-subject driving emotion is recognized: A multi-source transfer manifold learning model

• Published in: Biomedical signal processing and control Vol. 112; p. 108454
• Main Authors: Zhang, Wenqi, Qin, Yanjun, Zhang, Shanghang, Tao, Xiaoming
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.02.2026
• Subjects: Cross-subject; Driving emotion; Electroencephalogram (EEG); Manifold learning; Multi-source transfer learning; Multi-source transfer learning; Electroencephalogram (EEG); Manifold learning; Driving emotion; Cross-subject
• ISSN: 1746-8094
• DOI: 10.1016/j.bspc.2025.108454

The emotional state of drivers is closely related to driving safety and can be monitored through human body data. However, facial-based research is often affected by factors such as lighting and occlusions, leading many studies to shift towards physiological data. Moreover, work based on electroencephalogram (EEG) faces challenges with low accuracy in cross-subject detection. To address these issues, this paper investigates the driving emotions using multi-source transfer manifold learning based on EEG. We propose a manifold learning model that uses symmetric positive definite (SPD) matrix manifold features extraction and addresses the issue of low cross-subject validation accuracy through multi-source transfer learning. We use a dataset containing seven classes of physiological signals related to driving emotions and conduct experiments involving within-subject and cross-subject validations. Experimental results demonstrate that our model achieves high accuracy in detecting the emotional state of drivers, with an average accuracy of 94.28% in within-subject validation and 77.28% in cross-subject validation. This paper contributes to EEG-based emotion detection and driving safety. •Constructing a few-shot detection method for cross-subject emotion decoding based on multi-source transfer learning.•Incorporating a manifold learning structure based on SPD matrices in the EEG feature extraction process.•Studying seven categories of driving emotions and reaching high accuracy.