Speaker-independent speech emotion recognition on RAVDESS: MFCC, eGeMAPS and ComParE vs frozen WavLM, HuBERT, XLS-R and Whisper under nested leave-one-speaker-out evaluation, with end-to-end fine-tuning, protocol-inflation and contamination checks, and a per-clip comparison with human listeners.
reproducible-research scikit-learn pytorch opensmile whisper speech-processing affective-computing emotion-recognition linear-probing speech-emotion-recognition self-supervised-learning ravdess hubert wav2vec2 wavlm xls-r speech-foundation-models egemaps speaker-independent leave-one-speaker-out
-
Updated
Oct 9, 2026 - Python