Table 3.
Performance of replay voice spoofing detection models with fine-tuning.
| Dataset | System | With Pre-Training | Fine-Tuning Layer | Accuracy (%) | F1-Score (%) |
|---|---|---|---|---|---|
| ASVspoof2019 PA | Conventional | - | All layers | 92.94 | 86.92 |
| Proposed | 🗸 | FC | 88.6 | 85.58 | |
| 🗸 | Block 4 + FC | 93.7 | 92.67 | ||
| 🗸 | Block 3, 4 + FC | 96.2 | 93.97 | ||
| 🗸 | Block 2, 3, 4 + FC | 97.08 | 94.12 | ||
| 🗸 | Block 1, 2, 3, 4 + FC | 98.16 | 95.08 | ||
| 🗸 | All layers | 98.15 | 95.12 |