Affective brain-computer interfaces require models that can integrate EEG information across frequency bands and brain regions while remaining interpretable. This work proposes an explainable multi-frequency and multi-region fusion network (MFMR-FN) for affective BCI. The model represents EEG functional connectivity as symmetric positive definite matrices, learns discriminative Riemannian representations, and selects informative multi-region patterns for emotion recognition. Experiments show that the proposed framework outperforms existing methods and provides interpretable evidence about frequency-specific and region-specific brain connectivity patterns related to emotion and depression.
More details about this article are available at this link.