Dear Shahrzad Zandi, 

Try wavelet transform based thresholding of obtained coefficient according to artifacts frequency. 

can you upload your data?

I attached a sample. some channels have heart rate artifact which I want to extract from the muscle signal.

you can use this: (I attached data.txt and a sample result)

clear all

close all

clc

data=importdata('data.txt');

signal=data(:,5)'; % choose a channel, channel 5 for example

plot(signal)

signal_out=signal*0;

for i=100:100:size(data,1)-300

s=signal(i-99:i+100);

e_s=sum(s.^2);

if e_s< 50000 %% you can change these thresholds for different channels

signal_out(i-99:i+100)=s*.01+signal_out(i-99:i+100);

elseif e_s< 100000

signal_out(i-99:i+100)=s*.1+signal_out(i-99:i+100);

else

signal_out(i-99:i+100)=s*.95+signal_out(i-99:i+100);

end

end

figure, plot(signal_out,'r')

%% you can save signal_out 

Dear shahrzad

if you are able to work on simulink of matlab, using" filter design and analysis" wizard will help you to choose the best filter for filtering.

Dear Dr. Zandi,

I am of the conception that adaptive noise cancelling may be a reasonable alternative to be applied to your problem. To implement the foregoing approach you, however, need to record ECG signal (f ex. lead I) in a separate channel, while recording your EMG data. For more information please refer to

A. Cohen, Biomedical Signal Processing vol. I. Florida: CRC Press, Inc., 1986.

And MATLAB help with searching “adaptfilt.lms” command.

On the other hand, with a short glance to your data we do see that the amplitude of ECG artifacts is considerably lower than EMG amplitude; consequently, I personally prefer to simply neglect the artifacts instead of making the issue more complex.

In addition, I suppose there is a problem in channel 9 and 10 recordings.

Regards,