MS2FL: A modality-shared and modality-specific feature learning framework for multimodal emotion recognition

To facilitate reproducibility and further research in multimodal emotion recognition, we provide the processed versions of the SEED and SEED-V datasets used in this study. These datasets include preprocessed electroencephalography (EEG) and eye movement signals, as well as aligned labels and feature representations tailored to the proposed MS$^2$FL framework. While the original SEED and SEED-V datasets are publicly available, our processed versions are optimized for the tasks explored in this paper. Researchers interested in using the data are required to obtain permission from the SEED project website and may contact the authors for access to the corresponding processed versions.