CLG-MSTS: Contrastive learning-guided multi-scale temporal-spatial network for cross-subject emotion recognition

Publication Type

Conference Proceeding Article

Publication Date

5-2026

Abstract

EEG-based deep learning methods have been widely applied to cross-subject emotion recognition. However, these methods fail to adequately account for individual differences among subjects, resulting in diminished model performance. To this end, we propose a novel method called Contrastive Learning-Guided Multi-Scale Temporal-Spatial Network (CLG-MSTS) to learn subject-invariant representations. Specifically, the method employs a temporal-spatial joint framework to capture multi-scale features, and incorporates an Efficient Channel-Temporal Attention (ECTA) module, futher enabling the model to focus attention on the most important features. Moreover, the model continuously guides the network to constantly optimize to minimize inter-subject variability by integrating supervised contrastive learning. Experiments on the DEAP and MAHNOB-HCI datasets indicate that our proposed method achieves state-of-the-art cross-subject emotion recognition performance and generalizes well to new subjects.

Discipline

Artificial Intelligence and Robotics

Research Areas

Intelligent Systems and Optimization

Publication

Proceedings of the 2026 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Barcelona, Spain, May 4-8

Identifier

10.1109/ICASSP55912.2026.11460512

Publisher

IEEE

City or Country

Los Alamitos, CA

Additional URL

https://doi.org/10.1109/ICASSP55912.2026.11460512

This document is currently not available here.

Share

COinS