Research Article
Detection of Snore from OSAHS Patients Based on Deep Learning
Table 5
Accuracy of three feature extraction methods and three models.
| Feature extraction method | Evaluation | 3 layers-CNN | 5 layers-CNN | LSTM |
| MFCC | Sensitivity | 0.83 | 0.82 | 0.84 | Specificity | 0.84 | 0.81 | 0.91 | Precision | 0.82 | 0.84 | 0.91 | Accuracy | 0.82 | 0.85 | 0.87 | F1-score | 0.82 | 0.83 | 0.87 |
| LPMFCC | Sensitivity | 0.58 | 0.6 | 0.67 | Specificity | 0.64 | 0.67 | 0.72 | Precision | 0.62 | 0.64 | 0.71 | Accuracy | 0.61 | 0.63 | 0.69 | F1-score | 0.6 | 0.62 | 0.69 |
| LPCC | Sensitivity | 0.54 | 0.56 | 0.56 | Specificity | 0.6 | 0.6 | 0.62 | Precision | 0.58 | 0.58 | 0.6 | Accuracy | 0.57 | 0.58 | 0.59 | F1-score | 0.56 | 0.57 | 0.58 |
|
|