Table 9 Classification results of normal and abnormal of the proposed methods and other literature in PCCD.
From: Multi-level feature encoding algorithm based on FBPSI for heart sound classification
Methods | Acc (%) | Spe (%) | Pre (%) | UAR (%) | F1 (%) |
|---|---|---|---|---|---|
PSD-CNN9 | 84.72 | 94.63 | – | 77.63 | – |
WST + CWT & 1D CNN + 2D CNN15 | 98.57 | 98.00 | 98.00 | 98.00 | 98.00 |
DNN + stem_block + attention + DWconv16 | 97.70 | 92.22 | 98.98 | 98.20 | 98.59 |
42 MFCCs-KNN17 | 95.78 | – | – | – | 96.00 |
12 statistical features-SVM18 | 97.78 | 97.78 | – | 97.78 | – |
CTENN21 | 96.40 | 97.40 | – | 92.87 | 95.16 |
Conv-Encoder22 | 97.90 | 96.50 | 98.80 | 98.40 | – |
PIFS-ResNet1823 | 85.00 | – | – | – | 86.00 |
Our method—FBPSI + MLFE-EBTC | 98.73 | 95.65 | 98.88 | 99.53 | 99.21 |
