Table 5.
Performance comparison of various studies of Task I.
| Refs. | Physionet challenge ranking | Method | WAcc (train) | F-score (train) | WAcc (test) | F-score (test) |
|---|---|---|---|---|---|---|
| Reference 26 | 1 | Mel-spectrogram and lightweight CNN | 0.804 | 0.647 | 0.747 | 0.622 |
| Reference 27 | 2 | Parallel hidden semi-Markov models | 0.817 | 0.645 | 0.758 | 0.625 |
| Reference 28 | 3 | Hierarchical multi-scale convolutional network | 0.836 | 0.714 | 0.768 | 0.699 |
| Reference 44 | 4 | Hybrid Network | – | – | 0.768 | 0.731 |
| Reference 45 | 5 | Multiple instance learning framework | 0.828 | 0.568 | 0.734 | 0.544 |
| This study | – | Deep features + RFE + RF | 0.96 | 0.94 | 0.93 | 0.91 |