How to filter a sensor-measured signal to its theoretical form

Hello,

I used sensors to measure EMG signals, and it looks like this: (the original data is attached)

However, its theoretical form is this:

That is, it seems the lab measured data is subjected to some bias. How can I remove that bias?

Thank you very much if you could answer me this question!

Answer 1

[Star Strider]

Using the results of the fft, a highpass cutoff frequency of ‘0.02*Fs’ appears to be appropriate —

T1 = readtable('https://www.mathworks.com/matlabcentral/answers/uploaded_files/984450/1.txt', 'VariableNamingRule','preserve')

T1 = 2035×2 table
    Var1     Var2 
    _____    _____

    75954    80997
    74125    80310
    76792    85030
    76799    85062
    76807    85087
    76821    85126
    76838    85170
    76873    85240
    76902    85318
    76944    85411
    76993    85506
    77036    85619
    77086    85720
    77128    85833
    77166    85943
    77203    86043

s1 = T1.Var1;
s2 = T1.Var2;

Fs = 1;                                                     % Sampling Frequency (Use The Correct Value)
L = size(T1,1);
t = linspace(0, L-1, L)/Fs;                                 % Time Vector

figure
plot(t, s1,    t, s2)
grid
legend('s_1', 's_2', 'Location','best')

Fn = Fs/2;                                                      % Nyquist Frequency
NFFT = 2^nextpow2(L);                                           % Efficient 'fft' Length
FTs1s2 = fft([s1 s2]-mean([s1 s2]),NFFT)/L;                     % Fourier Transform (Subtract 'mean' To Show Peaks More Clearly)
Fv = linspace(0, 1, NFFT/2-1)*Fn;                               % Frequency Vector
Iv = 1:numel(Fv);                                               % Index Vector

figure
plot(Fv, abs(FTs1s2(Iv,:))*2)
grid
xlabel('Frequency')
ylabel('Amplitude')
legend('s_1', 's_2', 'Location','best')
xlim([0 0.15*Fs])

s1s2_filt = highpass([s1 s2], 0.02*Fs, Fs, 'ImpulseResponse','iir');             

figure
subplot(2,1,1)
plot(t, s1s2_filt(:,1))
grid
% xlabel('Time')
ylabel('Amplitude')
title('s_1')
subplot(2,1,2)
plot(t, s1s2_filt(:,2))
grid
xlabel('Time')
ylabel('Amplitude')
title('s_2')

.

[yuh zhang]

Hello! You are awsome, thank you so much for helping me!

But, in this code:

s1s2_filt = highpass([s1 s2], 0.02*Fs, Fs, 'ImpulseResponse','iir');  

is "0.02*Fs" chosen because there is a power peek at fre=0.02 in frequency-Amp chart?

[Star Strider]

I very much appreciate your compliment!

As always, my pleasure!

That frequency was chosen because it includes the first peak at about 0.025*Fs and everything greater than that. (There is no significant high-frequency noise, so a bandpass filter is not necessary with this signal.) You can experiment with other frequencies (between 0.01*Fs and 0.02*Fs) to see if that produces any better results.

It would be straightforward to also experiment with a bandpass filter. The upper frequency should likely be in the range of 0.04*Fs to 0.1*Fs for best results, depending on the result you want. This would eliminate what little high-frequency noise that exists in the signal, and depending on the chosen frequency, will also elimiinate all or part of the other two peaks (referring to the fft plot).

.