Refs. | Method | Performance | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Signal preprocessing | Signal generation | Number of signals | FE | FS | Classification method | Validation method | Number of selected features | Classifier | Testing data | Average time consumed for 50 consecutive segments | |||
Ac (%) | Se (%) | Sp (%) | Â | ||||||||||
[10] | Mean subtraction, Moving average, High-pass, Low-pass Butterworth filters | NA | 1 | TCSC | NA | Threshold based SAA | NA | 1 | Â | 98.1 | 80.9 | 98.5 | NA |
[44] | Moving average, High-pass, Low-pass Butterworth filters | Augmented ECG with MVMD | 2 | 54 features with conventional methods | GA and SFFS | ML based SAA | Five-fold CV | 20 | SVM | 99.0 | 97.4 | 99.2 | 4.7Â s |
KNN | 98.9 | 97.3 | 99.1 | ||||||||||
Boosting | 98.4 | 97.4 | 98.5 | ||||||||||
[8] | Moving average, High-pass, Low-pass Butterworth filters | Fully augmented ECG with MVMD | 3 | 100 deep features with CNN | NA | DL based SAA | Five-fold CV | 100 | SVM | 98.8 | 94.9 | 99.5 | Â |
KNN | 99.1 | 97.2 | 99.2 | Â | |||||||||
Boosting | 99.3 | 97.1 | 99.4 | 7.0Â s | |||||||||
Proposed algorithm | Moving average, High-pass, Low-pass Butterworth filters | Fully augmented ECG with MVMD | 3 | 93 features with conventional methods | SFFS | ML based SAA | Five-fold CV | 36 | SVM | 99.6 | 98.2 | 99.8 | 7.9Â s |
KNN | 99.3 | 97.7 | 99.9 | ||||||||||
Boosting | 98.2 | 95.4 | 98.9 |