Cardiac Arrhythmia Detection using Teager Energy Operator and Shallow Neural Networks

Cardiac Arrhythmias can be fatal if not identified in time. Prior identification of these using Electrocardiogram(ECG) signals has been a very important part of research in biomedicine. However, mostly ECG analysis methods are restricted to conventional offline PC host operation. The author here presents an arrhythmia classification method implemented on a Digital Signal Processing (DSP) using MATLAB to classify six heartbeat conditions: “ normal sinus rhythm (N), atrial fibrillation (AFIB), left bundle branch block (LBBB), right bundle branch block (RBBB), Atrial Flutter (AF) and Ventricular Flutter (VF) ”. The algorithm uses an IIR Butterworth digital filter to normalise the input signals. The classification task is performed using a shallow neural network and various other machine learning algorithms. 17 ECG records are obtained from the MIT_BIH dataset in the PhysioNet repository as the input data. The data obtained from the above dataset was pre-processed through MATLAB from which multiple important features were extracted. The features were then passed to various machine learning algorithms and a shallow neural network for classification into the aforementioned six arrhythmias. The findings show that the described model works well for the input dataset.