学术文献库

The human heart is a complex system exhibiting stochastic nature, as reflected in electrocardiogram (ECG) signals. ECG signal is a weak, non-stationary, and nonlinear signal, which indicates the health of a heart in terms of temporal variations of electromagnetic pulses from the heart. Abnormal fluctuations in ECG signal invokes the possibility of various cardiovascular disorders, which is diagnosed through intuitive analysis of the ECG reports by the medical practitioners. This could be made fast, accurate, and simple by imposing advanced nonlinear tools on the recorded ECG signals. In this paper, a well-known nonlinear technique, i.e., Empirical Mode Decomposition (EMD) method is adopted to extract the hidden information in the recorded ECG signal. Here, we try to explore the human heart as a dynamic model and perform EMD on ECG reports distinguishing arrhythmia from normal data obtained from the widely used MIT-BIH database. EMD essentially involves the decomposition of the signal into a finite number of Intrinsic Mode Functions (IMFs), keeping its original properties unaltered. For analysis, we use the powerful Savitzky-Golay (SG) filter for removing non-stationary noises from the ECG signals. The popular nonlinear parameter Hurst Exponent (H) is estimated for every IMF by R/S technique. We identified a distinct margin of the H of 1st IMFs in between the normal and the arrhythmia affected patients. Our model confirms with 94.92% certainty the chances of occurrence of arrhythmia disease in patients by diagnosing ECG signals without performing other expensive and time-consuming techniques such as Holter test, echocardiogram, and stress test.