Abstract
In this paper, the heart rate variability signals were utilized to discriminate different cardiac abnormalities. The performance of traditional feature extraction techniques and new complexity estimators was compared using hidden Markov model (HMM). This was done by analyzing data recorded for healthy subjects and patients suffering from congestive heart failure (CHF) and myocardial infarction (MI) diseases. The techniques utilized included time, frequency parameters as well as approximate entropy, sample entropy, detrended fluctuation analysis coefficient, fractal dimension, Poincare plot and largest Lyapunov exponent. Results have shown that using complexity-based estimators gives higher rates for classifying cardiac abnormalities. Classification rate reaches to 98.18%.