A benchmark across 19 datasets shows contrastive pretraining and handcrafted features give the strongest generalization for EDA and PPG emotion recognition.
0_back" were considered low in cognitive demand and thus assigned low arousal and positive 20 valence. Conversely, tasks labeled as
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.HC 1years
2026 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
FEEL: Quantifying Heterogeneity in Physiological Signals for Generalizable Emotion Recognition
A benchmark across 19 datasets shows contrastive pretraining and handcrafted features give the strongest generalization for EDA and PPG emotion recognition.