EmoMind: Decoding Affective Captions from Human Brain fMRI

EmoMind introduces an innovative pipeline for generating affective captions directly from fMRI signals, combining semantic scene descriptions with a continuous 34-dimensional emotion vector. It employs classifier-free guidance to balance content preservation and emotional expressivity, outperforming GPT-4 in generating individualized affective captions across multiple evaluation axes. This advancement enhances the understanding of individual affective brain organization and offers a new methodology for integrating emotional context into AI-generated text, which is crucial for practitioners working on brain-computer interfaces and affective computing.