Dynamic Gaussian filter for muscle noise reduction in ECG signal

We propose a dynamic filter with variable smoothing effect in different parts of the signal in order to suppress the EMG noise significantly outside the QRS region while applying much less smoothing to the QRS complex. Gaussian filtering technique is used for this purpose where the width of the Gaussian filter changes for each sample of the signal according to the slope of the signal at that point. We estimate the slope of the signal similar to the techniques used in QRS detection algorithm. The proposed filter is used as a preprocessing stage in ecgpuwave algorithm available on Physionet and Saadat Arrhythmia Detection Algorithm. We evaluate the performance of this filter in signal denoising, and preserving the morphology on MIT-BIH Arrhythmia and CSE databases and compare it with 35Hz and 25Hz low-pass filters normally used for muscle noise reduction. The results show that the proposed filter performs better than 35Hz and 25Hz low-pass filters in removing the EMG noise while the delineation, QRS detection and classification performance on CSE and MIT-BIH databases remain intact or even improved.