Paper ties singing emotion recognition to interpretable vocal technique cues
Original title: Interpretable Singing Emotion Recognition Using Acoustic and Vocal Technique Features
SHAP and Grad-CAM link predictions to vibrato and pharyngeal technique, while singing-specific features reduce cross-singer variability.
The authors propose a singing emotion recognizer built from a compact set of singing-specific descriptors covering prosody, spectral features, phonetics, and vocal technique. SHAP explains model decisions, and Grad-CAM inspects learned representations of vocal techniques. They report lower cross-singer variability and, on out-of-domain VocalSet evaluation, associations between emotional expression and ground-truth technical execution.
- Vibrato is described as context-dependent on articulation rate.
- Out-of-domain evaluation uses the VocalSet corpus.
- The framework treats the singing voice as distinct from speech.
From the original article
1 min read at arxiv.orgSinging emotion recognition relies predominantly on deep learning models that achieve strong performance but offer limited insight into the vocal characteristics driving emotion, of- ten overlooking the singing voice as a distinct signal.
By Fotini Kiriakopoulou · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.