Table 3 Comparison of our work with related CNN-based methods with the test set
Study | Classifiers | F1(AF) | F1(Normal) | F1(Other) | F1(Noisy) |
|---|---|---|---|---|---|
Hsieh [27] | 1D CNN | 80.80 | 90.40 | 66.20 | 75.30 |
Xiong [26] | 1D CNN | 85.80 | 91.90 | 81.60 | - |
Jiang [28] | 1D ResNet + Bi-LSTM | 88.00 (weight average) | |||
Zhao [29] | 2D Kalman CNN | 79.18 | 89.29 | 72.25 | 52.50 |
Fang [12] | 2D Dual-Vgg16 | 83.00 | 90.00 | 75.00 | 83.00 |
Lee [13] | 2D ECM Bit-CNN | 89.73 | 81.06 | 74.45 | 62.22 |
Our work | 2D MsCWT CNN | 91.22 | 93.75 | 86.96 | 85.71 |