A multimodal dataset for emotion and personality recognition in real-world tasks using egocentric vision and physiological signals.

Multi-dataset joint pre-training with covariance centroid alignment enables cross-dataset generalization of EEG-based emotion recognition.