Arrhythmia Detection Based on the Reduced Features with K-SVD Sparse Coding Algorithm

Arrhythmia is a subnormal heartbeat that causes Tachycardia or Bradycardia. ECG signal is extensively used for the analysis of heart diseases. The ECG is used to measure the rate and regularity of heartbeats. With the help of the study of the ECG signal, the presence of any damage to the heart can be identified. Arrhythmia classification is an important part of the research of medical images and signal processing. This paper aims to apply the K-SVD technique so that both the best features selected, and the accuracy of arrhythmia classification are improved. Additionally, the required space for storage has been reduced. For each ECG, we extracted 100 features of each ECG signal. Afterward, we applied statistical descriptors for extracted features which include Mean, Energy, Variance, Covariance, and Standard Deviation (S.D). The next step is the training process. We applied MLP neural networks. We have used two kinds of datasets in our research paper, UCI Machine Learning Repository and the MIT-BIH dataset. The accuracy rate of the proposed method is 98.90% and 99.14% for the UCI and MIT-BIH Dataset, respectively.