To develop an automated emotion recognition system that captures the nuances of spontaneous affect using a large corpus of authentic facial emotion expressions.
Key Findings:
Model predictions closely matched human annotations with a mean z-score of 1.97.
Strong recognition of joy, sadness, and fear was confirmed through external evaluation against acted datasets.
Interpretation:
Limitations:
The study may be limited by the specific context of psychotherapy sessions and may not generalize to other settings.
The reliance on human annotations could introduce subjectivity in the training data.
