Machine Learning based Cardiac Arrhythmia detection from ECG signal

This paper presents the analysis of heart diseases that are categorized as arrhythmia based on Electrocardiogram (ECG). ECG database of different disease conditions was analyzed. The ECG signals are filtered to remove noise which is caused due to powerline interface or Electromyogram. This filtered signal is segmented to smaller pieces of ECG so that feature extraction is accurate. The features extracted are Peak to peak Interval (R-R Interval), BPM (Beats per minute), P wave to QRS peak. The data set is classified using an SVM classifier algorithm. This algorithm classifies the input ECG signal with varying feature parameters to two different types of arrhythmia. This approach has achieved an accuracy of 91% and the performance regarding other criteria such as precision, recall and F1 score were significantly better indicating the success of the proposed method.