ECG beat classification using proposed pattern adaptive wavelet-based hybrid classifiers

In this paper, we have proposed a Pattern adaptive wavelet-based hybrid approaches for classification of arrhythmia beats. The main aim is to categorize and discriminate Electrocardiogram (ECG) beats into normal heart beat (Nb) and abnormal beats like atrial premature contraction beat (Ab), premature ventricular contraction beat (Vb), left bundle branch block beat (Lb), paced beat (Pb) and right bundle branch block beat (Rb). The features are extracted using our new designed wavelet, i.e., Pattern adaptive wavelet and the existing symlet4 wavelet. Various machine learning classification methods like K-Nearest Neighbors, Random Forest, Decision Tree, Bagged Decision Tree, Extreme Learning Machine and Naive Bays (NB) are implemented and performance of each classifier is evaluated. We have proposed two Hybrid Classifiers using ensemble learning techniques to improve the performance. The hybrid approach has been tested with Massachusetts Institute of Technology-Boston's Beth Israel Hospital Arrhythmia Database ECG records. The result shows that the proposed Pattern adaptive wavelet-based hybrid approaches outperform the individual classifiers with increased accuracies of 99.16% and 97.8% using Pattern adaptive wavelet features. The Hybrid Classifier-I shows 0.64% and Hybrid Classifier-II shows 2.6% accuracy more than the individual classifier which showed the highest among the base classifiers.