Effective Removal of Baseline Wander from ECG Signals: A Comparative Study

Electrocardiogram (ECG) signal classification is an essential task to diagnose arrhythmia clinically. For effective ECG analyses, it has to be decluttered from embedded low and high frequency noise. Low frequency noise include baseline wander and high frequency noise include power line interference. We provide a comparative study for the task of baseline wander removal from ECG signals using different variants of Empirical Mode Decomposition, Median Filtering and Mean Median Filtering with a major emphasis on variational mode decomposition as it is a relatively new technique and much more robust towards noise. The comparison between the aforementioned techniques depicted that variational mode decomposition estimates better baseline as compared to other techniques in terms of pearson correlation, percentage root mean square difference and maximum absolute error. However, the time required to decompose the signal is relatively higher than the filtering techniques.