Emotion Classification on RAVDESS four-class subset n=672
71AccuracyMFCC + LogReg (n=20)
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| MFCC + LogReg (n=20)Feat.=40-dim, Evaluation Protocol=stratified five-fold cross-validation2026.06 | 71 | 69 | 68.6 | 68.8 | 70.8 | 70.3 | 82.8 | 67.7 | 54.3 | |
| MFCC + LogReg (n=40)Feat.=80-dim, Evaluation Protocol=stratified five-fold cross-validation2026.06 | 69 | 68.6 | 68.5 | 68.5 | 69.2 | 64.6 | 80 | 65.3 | 64.2 | |
| ESN + Ridge (n_res=500)Feat.=1000-dim, Evaluation Protocol=stratified five-fold cross-validation2026.06 | 68.3 | 67.3 | 68.1 | 67.6 | 68.3 | 70.3 | 76.8 | 60.7 | 62.4 | |
| MFCC + LogReg (n=13)Feat.=26-dim, Evaluation Protocol=stratified five-fold cross-validation2026.06 | 66.8 | 65 | 64.3 | 64.6 | 66.6 | 66.5 | 81.8 | 59.9 | 50.3 | |
| wav2vec2-large-superb-erFeat.=1024-dim, Evaluation Protocol=zero-shot2026.06 | 48.8 | 64.2 | 45.8 | 39.4 | 40 | 49.1 | 72.6 | 1 | 34.8 | |
| wav2vec2-base-superb-erFeat.=768-dim, Evaluation Protocol=zero-shot2026.06 | 43.3 | 54.4 | 45.6 | 36.6 | 35.1 | 23.6 | 61.8 | 13.6 | 47.4 | |
| Majority-classFeat.=n/a2026.06 | 28.6 | 7.1 | 25 | 11.1 | 12.7 | 44.4 | 0 | 0 | 0 | |
| Random (uniform)Feat.=n/a2026.06 | 24.9 | 24.9 | 24.3 | 24.2 | 25.5 | 30.7 | 26.8 | 24.2 | 15 |