Morphologic Based Feature Extraction for Arrhythmia Beat Detection

Abstract

Heart disease is one of the diseases which has highest mortality rate recently. Heart's electrical activity examination and interpretation are very important for the understanding of diseases. In this study, electrocardiogram signals are analyzed, then patient's healthy and arrhythmia beats are extracted. RR, QRS, Skewness and Linear Predictive Coding coefficients of the signals are considered for classification of the data. K-NN, Random SubSpaces, Naive Bayes and K-Star classifiers are used. The highest accuracy is obtained with the K-NN algorithm (98.32%). At the second stage of the K-NN algorithm, accuracy levels are examined by changing the 'k' parameter.