The paper introduces an olfactory-emotion dataset involving 111 participants, with synchronized EEG, ECG, and PPG recordings and odor labels in a two-dimensional arousal-valence space. Its proposed STF-HFNet combines adaptive frequency aggregation, reciprocal guided attention, and hybrid spatial-channel fusion to address non-stationarity, latency differences, and cross-modal heterogeneity. According to the abstract, the model reaches 88.34% accuracy on AMIGOS and 92.40% on the authors’ dataset, exceeding the compared state of the art by 8.27% and 5.07%, respectively.
No heat snapshots are available in the last 24 hours.