Assessment of Values of Time-Domain and Frequency-Domain Parameters for ECG Signals Through HRV Analysis Using Symlets for Arrhythmia Prediction

摘要

In this paper, we present a work on HRV analysis of ECG signals using three time-domain parameters namely SD ratio, pNN50 and RMSSD and one frequency-domain parameter namely LF/HF ratio. For this work we obtained the ECG signal data from the MIT-BIH database. A total of 40 ECG signals, 10 each for normal sinus rhythm, for atrial fibrillation, for supra-ventricular fibrillation and for premature ventricular contraction, were used for experimentation. All ECG signals are of 30 min duration. The R-peak was detected using the Symlet5 wavelet at second level of decomposition. The accuracy of R-peak detection was found to be 97 %. Through R-peak detection and through the determination of the RR intervals, we estimated the values of mean and of standard deviation of the 4 parameters. Using the information from RR intervals, we also obtained the Poincaré plots and the power spectral density plots for the 40 signals. Based on the obtained values of the parameters we comment on the nature of values of the parameters in the paper, for normal and abnormal conditions. We further observe that by visual inspection also of the Poincaré plot and of the power spectral density plot of an unclassified ECG signal, the signal can be classified as normal or abnormal.