Table 4 Performance metrics for different methods and datasets.
From: Speech emotion recognition with light weight deep neural ensemble model using hand crafted features
Method | Dataset | Accuracy | Weighted F1 | AUC-ROC | AUC-PRC |
|---|---|---|---|---|---|
1D CNN | EmoDB | 98.60 | 98.06 | 100.00 | 100.00 |
CNN_Bi-LSTM | EmoDB | 94.39 | 93.74 | 99.87 | 99.39 |
Ensemble | EmoDB | 98.13 | 98.12 | 100.00 | 99.95 |
1D CNN | RAVDESS | 96.18 | 96.22 | 99.91 | 99.50 |
CNN_Bi-LSTM | RAVDESS | 97.57 | 97.29 | 99.92 | 99.57 |
Ensemble | RAVDESS | 97.57 | 97.56 | 99.95 | 99.68 |
1D CNN | TESS | 100.00 | 100.00 | 100.00 | 100.00 |
CNN_Bi-LSTM | TESS | 99.82 | 100.00 | 100.00 | 100.00 |
Ensemble | TESS | 100.00 | 100.00 | 100.00 | 100.00 |
1D CNN | SAVEE | 96.88 | 97.63 | 99.98 | 99.87 |
CNN_Bi-LSTM | SAVEE | 89.58 | 87.42 | 99.28 | 96.31 |
Ensemble | SAVEE | 98.44 | 98.45 | 99.98 | 99.88 |
1D CNN | CREMA-D | 96.07 | 96.24 | 99.77 | 99.29 |
CNN_Bi-LSTM | CREMA-D | 97.31 | 97.34 | 99.81 | 99.50 |
Ensemble | CREMA-D | 98.66 | 98.65 | 99.88 | 99.72 |
