Cross-Subject and Cross-Session EEG Emotion Recognition based on Multi-Source Structural Deep Clustering

• Published in: IEEE transactions on cognitive and developmental systems pp. 1 - 15
• Main Authors: Chen, Yiyuan, Xu, Xiaodong, Qin, Xiaowei
• Format: Journal Article
• Language: English
• Published: IEEE 24.02.2025
• Subjects: Brain modeling; Convolutional neural networks; Cross-Session; Cross-Subject; Data mining; EEG Emotion Recognition; Electroencephalography; Emotion recognition; Feature extraction; Long short term memory; Multi-Source; Nearest neighbor methods; Structural Deep Clustering; Support vector machines; Training
• ISSN: 2379-8920; 2379-8939
• DOI: 10.1109/TCDS.2025.3545666

Individual fluctuations and temporal variabilities of Electroencephalogram (EEG) pose challenges in precisely identifying emotions. Although a model performs well with data specific to a certain subject or session, the fluctuations in EEG data can significantly impair the performance on a different subject or session. As a result, current approaches synchronize the original and new subject or session feature distributions. Directly matching EEG data across individuals or sessions may undermine the inherent distinguishability due to the heterogeneity in data distribution. Instead of direct alignment, this work utilizes multi-source structural deep clustering to identify the inherent structural knowledge of the target itself and regularize it through the distribution of source labels. Furthermore, the method was implemented on the intermediate output utilizing high-confidence features to improve pattern identification in the latent feature space. This led to more distinct differentiations across subdomains with varying labels. Comparative analyses were performed with state-of-the-art (SOTA) models on SEED and SEED-IV datasets. The model proposed outperformed other baseline models under the strict leave-one-subject-out strategy, reaching an average accuracy of 88.20%/90.06% in a cross-subject/cross-session experiment on SEED and 71.49%/69.96% in SEED-IV. This research provides a novel approach to align EEG features without the need for direct distance calculation.