Table 6 Results (unweighted average recall (UAR)) with all features of proposed acoustic features set (EmoFt 200 features) and the INTERSPEECH 2013 ComParE feature set (6373 features) for ternary arousal and valence tasks; normalisation on training fold (mean/variance); leave-one-singer-out cross validation
[UAR %] | EmoFt | ComParE | Chance | ||
|---|---|---|---|---|---|
SVM complexity C | 0.1 | 1.0 | 0.1 | 1.0 | - |
Global (training fold) feature normalisation: | |||||
Eleven emotion classes | 22.4 | 24.4 | 28.0 | 28.0 | 9.1 |
Valence (three classes) | 36.4 | 40.2 | 46.2 | 46.2 | 33.3 |
Arousal (three classes) | 57.9 | 52.4 | 54.6 | 54.6 | 33.3 |
Per singer feature normalisation: | |||||
Eleven emotion classes | 23.6 | 28.1 | 38.2 | 38.2 | 9.1 |
Valence (three classes) | 40.1 | 43.6 | 48.7 | 48.7 | 33.3 |
Arousal (three classes) | 57.6 | 49.9 | 57.6 | 57.6 | 33.3 |