ECG beat classification based on signal decomposition: a comparative study

Two effective ECG beat classification methods based on signal decomposition were compared in terms of effective feature selection and noise tolerance. The HOS-DWT-FFBNN method associated with the linear correlation based filter (LCBF) provides imposing capability to select the more representative features than the IC reordering method OWSL associated with the ICA-SVM method. Both methods are insensitive to the stationary artifacts, power line interference and white Gaussian noise. The HOS-DWT-FFBNN method is especially tolerant to non-stationary artifacts, baseline wander and muscle artifact, when compared to ICA-SVM. More than 90% accuracy can be retained with the HOS-DWT-FFBNN method when the SNR decreases to 10 dB, while only 63.16% and 78% accuracies, respectively, can be reserved with the ICA-SVM method.