Research Article
Semisupervised Deep Features of Time-Frequency Maps for Multimodal Emotion Recognition
Table 6
Confusion matrix of the four-class scenario.
| ā | Predicted emotion | Sens | Prec | Kappa | F1 score | Happy | Angry | Calm | Sad |
| Actual emotion | Happy | 0.933 | 0.021 | 0.028 | 0.018 | 0.933 | 0.931 | 0.862 | 0.932 | Angry | 0.019 | 0.937 | 0.018 | 0.026 | 0.937 | 0.922 | 0.844 | 0.929 | Calm | 0.032 | 0.02 | 0.923 | 0.025 | 0.923 | 0.928 | 0.857 | 0.926 | Sad | 0.018 | 0.038 | 0.025 | 0.919 | 0.919 | 0.931 | 0.861 | 0.925 |
|
|