NOVEL METHOD TO FIND THE PARAMETER FOR NOISE REMOVAL FROM MULTI-CHANNEL ECG WAVEFORMS

Abstract

In general, electrocardiogram (ECG) waveforms are aby noise and artifacts and it is essential to remove the noise in order to support any decision making for specialist. It is v ery difficult to remove the noise from 12 channel E CG waveforms using standard noise removal methodologies. Removal of the noise f rom ECG waveforms is majorly classified into two ty pes in signal processing namely Digital filters and Analog filters. Digital filters are more accurate than analog filters becau se analog filters introduce nonlinear phase shift. Most advanced research digit al filters are FIR and IIR.FIR filters are stable a s they have non-recursive structure. They give the exact linear phase and eff iciently realizable in hardware. The filter respons e is finite duration. Thus noise removal using FIR digital filter is better option iwith IIR digital filter. But it is very difficult to find the cut-off frequency parameter for dynamic multi-channel ECG waveforms u sing existing traditional methods. So, in this research, newly introduced Multi-Swarm Optimization (MSO) methodology for automatically id entifying the cut-off frequency parameter of multichannel ECG waveforms f or low-pass filtering is inspecting. Generally, the spectrums of the ECG waveforms are extracted from four classes: normal srhythm, atria fibrillation, arrhythmia and sup raventricular. Baseline wander is removed using the Moving Median Filter. A datase t of the extracted features of the ECG spectrums is used to train the MSO. The performance of the MSO with various parameters is i nvestigated. Finally, the MSO-identified cut-off fr equency parameter, it’s applied to a Finite Impulse Response (FIR) filter. The resu lting signal is evaluated against the original clea n and conventional filtered ECG signal.