A novel method for R-peak detection in noisy ECG signals using EEMD and ICA

In this paper, we present a novel algorithm for R-peak detection in noisy ECG signals using ensemble empirical mode decomposition (EEMD) and independent component analysis (ICA). First, we decompose the ECG signal into several Intrinsic Mode Functions (IMFs). Preprocessing is performed to select the number of IMFs (i.e., 8). These IMFs are used as the mixture signals for the ICA algorithm. Since the QRS complex is usually the strongest component in the ECG signal, if one independent component (IC) is only separated using ICA, it most probably contains the QRS complex. Pan-Tompkins algorithm is then applied to the separated IC for R-Peak detection. The QT database was used for the performance evaluation of the algorithm in different noise levels. The results show an average mean squared error (MSE) of 3.025 samples (12.1 ms), an average mean absolute error of 1.22 samples (4.87 ms), and an average MSE variance of 0.61 samples (2.44 ms) in 3 different noise levels compared with the manually annotated beats.