Investigating the differences between atrial fibrillation and normal ECG signals using the EMD method

this paper presents the application of empirical mode decomposition (EMD) for analyzing electrocardiograph (ECG) signals and detecting Atrial Fibrillation (AF) cases. EMD generates a limited and small number of Intrinsic Mode Functions (IMFs). The decomposition is based on the direct extraction of components related to various intrinsic time scales. In this paper, analytic functions corresponding to IMFs are obtained using the Hilbert transform. Using a Central Tendency Measure (CTM), the radius of a circle in which 95% of analytic function points are located within the circle is determined. ECG signals like other biological signals are generally non-stationary. Although traditional processing tools such as short-time Fourier or wavelet transforms are applied for studying biological signals, they are unable to probe completely the ECG signals. EMD is used as a proper technique for analyzing and decomposing non-stationary data. The area of this circle in the complex plane has been used as a feature in order to categorize normal ECG signals from the AF ECG signals. Our results indicate that the area criterion of IMFs in the complex plane in the normal cases are greater than those of the AF cases. In this way, the proposed method can be effective in differentiating the ECG signals of normal subjects and cases with AF.