Exploring the Non-Local Similarity Present in Variational Mode Functions for Effective ECG Denoising

Abstract

ECG is corrupted by various noises and denoising of ECG is central to proper diagnosis of cardiac diseases. The main objective of this work is to exploit the efficacy of non-local means (NLM) for ECG denoising. In the presented work, noisy ECG signal is decomposed into variational mode functions (VMFs) using variational mode decomposition (VMD) technique. The decomposed VMFs represents the different frequency band of the noisy ECG signal. The non-local similarity present in each VMFs were exploited using NLM estimation for effective ECG denoising. The two-stage VMD decomposition and NLM estimation process is performed on different set of VMFs at both stages. The proposed method is tested upon MIT-BIH Arrhythmia database. The denoising performance of existing techniques is compared to proposed method. The computed results shows that the proposed method gives superior denoising performance.