| 算法 | IEMOCAP | Emo-DB |
| CNN | 75.80% | 76.58% |
| LSTM | 76.10% | 83.73% |
| RNN | 76.22% | 85.91% |
| CNN + LSTM | 78.17% | 89.79% |
| Bi-GRU | 77.01% | 89.22% |
| Self-Attention | 77.87% | 88.91% |
| Bi-GRU + Self-Attention | 79.21% | 90.58% |
| Multi-head Attention | 78.13% | 89.37% |
| 本文算法 | 81.04% | 94.93% |