Signal Denoising Technique Employing Detrended Fluctuation Thresholding and Stationary Wavelet Transform in the Framework of Data-driven Multiresolution Analysis

This paper proposes a new approach for the signal denoising method based on empirical mode decomposition (EMD) and a novel metric based on detrended fluctuation analysis (DFA). The proposed denoising method first obtains data-driven signal representation called intrinsic mode functions (IMFs) using EMD. Next, DFA is used to threshold the predominantly noisy modes based on their scaling exponent and higher frequency content. Finally, using a shrinkage method on predominantly noisy modes, the denoised signal is reconstructed by the combination of the denoised modes and the remaining IMFs which were obtained as an output from the first step. A comprehensive simulation-based study demonstrates the superiority of the proposed method in comparison to the state-of-the-art methods in signal denoising. Furthermore, the proposed denoising scheme exhibits promising denoising results in practical applications for electrocardiogram (ECG) signals.