Classification of ECG Signal Using Machine Learning Techniques

Abstract

Electrocardiogram (ECG) signals are the impulses generated by the heart which are used to analyze the proper functioning of heart. Our work deals with the efficient analysis of Electrocardiogram (ECG) signals imported from MIT-BIH database into MATLAB platform, generation of the imported ECG signal, pre-processing the generated signal to remove the noises mainly the baseline wandering and power line interference from which features are extracted. For adequate study of the ECG signal Daubechies and Haar wavelet techniques are compared. Proper decomposition of the signal is achieved using Db4 and Db5 Daubechies wavelets as their scaling functions are analogous to ECG signal. PAN TOMPKINSONS algorithm is considered in our study as it serves best for precise identification of most prominent features namely QRS complexes, RR interval’s as they constitute the major data required for clinical analysis and research. After feature extraction ECG signals are trained using machine learning techniques for detecting the presence of Arrhythmia using different classifiers adopting Weka software.