Table 2 Performance evaluation of deep learning models
Metric | CNN6 | PaSST | ResNet38 | |||
|---|---|---|---|---|---|---|
| Â | Fbank | Mel spectrogram | Fbank | Mel spectrogram | Fbank | Mel spectrogram |
Accuracy (%) | 94.12 (± 0.1431) | 93.78 (± 0.1369) | 92.66 (± 0.3852) | 93.18 (± 0.2279) | 92.53 (± 0.6505) | 91.56 (± 0.7947) |
Specificity (%) | 97.3 (± 0.0009) | 97.22 (± 0.0011) | 96.75 (± 0.0016) | 97.02 (± 0.0014) | 96.8 (± 0.0016) | 96.04 (± 0.0042) |
Sensitivity (Recall, %) | 94.12 (± 0.1431) | 93.78 (± 0.1369) | 92.66 (± 0.3852) | 93.18 (± 0.2279) | 92.53 (± 0.6505) | 91.56 (± 0.7947) |
Precision (%) | 92.63 (± 0.2724) | 93.73 (± 0.5896) | 92.53 (± 0.5512) | 93.43 (± 0.1964) | 91.11(± 0.6216) | 92.81 (± 0.7480) |
F1 score (%) | 93.32 (± 0.1806) | 93.75 (± 0.3440) | 92.59 (± 0.4684) | 93.30 (± 0.1880) | 91.75 (± 0.5024) | 92.14 (± 0.6770) |