Table 5. Comparison of F1 scores between the unidirectional and bidirectional models.
| Inhalation | Exhalation | CASs | DASs | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Models | n of trainable parameters | F1 score | F1 score | F1 score | F1 score | ||||
| Segment Detection | Event Detection | Segment Detection | Event Detection | Segment Detection | Event Detection | Segment Detection | Event Detection | ||
| LSTM | 300,609 | 73.9% | 76.1% | 51.8% | 57.0% | 15.1% | 12.2% | 62.6% | 59.1% |
| SIMP BiLSTM | 235,073 | 77.8% | 84.1% | 55.8% | 62.4% | 19.8% | 17.9% | 68.8% | 68.9% |
| GRU | 227,265 | 76.2% | 78.9% | 59.8% | 65.6% | 24.6% | 20.1% | 65.9% | 62.5% |
| SIMP BiGRU | 178,113 | 80.1% | 86.1% | 63.7% | 70.0% | 25.0% | 22.2% | 70.3% | 71.3% |
| CNN-LSTM | 3,448,513 | 77.6% | 81.1% | 57.7% | 62.1% | 45.3% | 42.5% | 68.8% | 64.4% |
| SIMP CNN-BiLSTM | 3,382,977 | 80.0% | 85.8% | 60.4% | 66.2% | 50.8% | 50.2% | 70.2% | 70.2% |
| CNN-GRU | 2,605,249 | 78.4% | 82.0% | 57.2% | 62.0% | 51.5% | 49.8% | 68.0% | 64.6% |
| SIMP CNN-BiGRU | 2,556,097 | 80.1% | 85.9% | 62.4% | 68.4% | 52.6% | 51.5% | 69.9% | 69.5% |
The bold values indicate the higher F1 score between the compared pairs of models. SIMP means the number of trainable parameters is adjusted.