Major and Sub-Class Classification of Arrhythmia using Eigen Vectors in ConvNet

Cardiac Arrhythmia is a heart disease that corresponds to abnormal rhythm of heart. It means that the heart is either beating too quickly, too slowly, or sporadically. Arrhythmia is recognized and categorized effectively so as to improve the living conditions of the patients. The Electro Cardiogram (ECG) is a tool for recording electrical activity and determining the electrical impulses in the heart. There are four main classes of arrhythmia which occur due to abnormal heartbeat which are being classified. The main objective of this proposed work is to provide better performance in predicting arrhythmia since even a small error can become dangerous to a person's life. The existing methods uses CNN as the feature extraction model which delays the time of prediction. Here, a novel feature extraction method is introduced based on 1D-Convolutional Neural Networks using the Eigen Vectors functionality. This feature extraction model proves to outperform the existing works in accurately classifying the different classes of arrhythmia. Finally, the ANN model is trained using the K-fold Cross Validation method to achieve this performance and is compared with an ensemble model containing a SVM, ANN and a Decision Tree.