A unified signal processing and machine learning method for detection of abnormal heart beats using Electrocardiogram

In this paper, a unified signal processing and machine learning method to automatically process Electrocardiogram (ECG) signal for classification of heartbeat type is presented. The method is divided into three stages: signal processing and transformation, feature extraction, and classification. The method can classify a beat into one of eight classes. Thirty features are extracted from time and frequency domains of ECG signal. The data are obtained from MIT/BIH arrhythmia database. The classification results are found to have high accuracy of classification (99.73%). When compared to previously reported algorithms, the method exhibit great performance. The approach plays an important role in a decision support system for early detection of arrhythmias, which can greatly help in planning and timing of resuscitation.