Table 5 Comparison of different classification methods on the MIT-BIT arrhythmia database
From: Fast multi-scale feature fusion for ECG heartbeat classification
Authors | Features | Classifier | Classes | Average accuracy |
|---|---|---|---|---|
Manab Kumar Das et al. [26] | ST +WT + Temporal | MLP-NN | 5 | 97.5 % |
Inan et al. [27] | DyWT + timing information | Neural network | 2 | 95.16 % |
Jiang et al. [28] | Hermite transform coefficients + time interval | Block-based neural networks | 5 | 96.6 % |
Ince et al. [29] | Morphological-wavelet transform + PCA, temporal features | Optimal artificial neural networks | 5 | 95.58 % |
Martis et al. [30] | Bispectrum + PCA | SVM with RBF kernel | 5 | 93.48 % |
Proposed method | WPD + GND_ICA | SVM with RBF kernel | 4 | 97.54 % |