How to handle ECG muscle noise artifacts correctly ?

I'm trying to process ECG recordings taken by Holter (bipolar chest leads) during simulations in virtual reality. Therefore, people move a lot when measuring so I often encounter movement and muscle artifacts. What is the best approach or what filter design should be used to efficiently suppress the EMG artifacts which overlap the R waves. I will be happy for any advice. Thank you in advance (attachment is sample of used signal).

Answer 1

[Star Strider]

I would use a bandpass filter.

Example —

D = load('ecg.mat');
EKG = D.ecg;

L = numel(EKG);
Fs = 125;                                               % Use Actual Sampling Frequency
tv = linspace(0, L, L)/Fs;                              % Time Vector
Fn = Fs/2;

EKGmc = EKG-mean(EKG);
FTEKG = fft(EKGmc)/L;
Fv = linspace(0, 1, fix(L/2)+1)*Fn;
Iv = 1:numel(Fv);

figure
plot(Fv, abs(FTEKG(Iv))*2)
grid
xlim([0  1])
title('Fourier Transform')

EKGfilt = bandpass(EKG, [0.03  0.2]);

figure
subplot(2,1,1)
plot(tv, EKG)
grid
title('Unfiltered')
subplot(2,1,2)
plot(tv, EKGfilt)
grid
ylim([-0.5 1.5])
title('Bandpass Filtered')
xlabel('t')

Use the Fourier transform plot to guide the filter design.

[Spectro]

I was thinking mainly about the filters like Butterworth, Chebbyshev etc. because I know they exist in matlab. I didn't know about this one (bandpass();). It handled the signal quite well. Thank you for your quick response and I appreciate your help.

[Star Strider]

As always, my pleasure!

It is straightforward to design filters using command-line funcitons. I prefer elliptic filters (the ellip function and its firends) since it is computationally more efficent than the others.

That would go something like this:

Fs = 125;
Fn = Fs/2;
Wp = [0.03 0.2]/Fn;
Ws = Wp.*[0.95 1.05];
Rs = 50;
Rp =  1;
[n,Wn] = ellipord(Wp,Ws,Rp,Rs);
[z,p,k] = ellip(n,Rp,Rs,Wp);
[sos,g] = zp2sos(z,p,k);

figure
freqz(sos, 2^16, Fs)
set(subplot(2,1,1), 'XLim',[0 0.5])
set(subplot(2,1,2), 'XLim',[0 0.5])

EKGfilt = filtfilt(sos, g, EKG);

Use the corredct value for ‘Fs’. The rest automaticaly scales with it, so no other changes are necessary, unless the desired performance is different than in this filter.