How can I correct the baseline shift of the signal

I'm processing EMG signals an there seems to be a shift of the signal on the raw and filtered signals. I'm using a bandpass filter, and the correct cutt off frequencys. Any clue how can I fix it?

Answer 1

[Star Strider]

The baseline constant offset is by definition at 0 Hz, and since your bandpass filter eliminates everything below 0.2 times your Nyquist frequency, the baseline of your filtered signal will be 0, not the baseline of your original signal, since you filtered out the baseline offset. You can add it back in, however that is usually not worrthwhile. Just use the filtered signal as it exists.

Also, use the phase-neutral filtfilt function, not filter, to do the actual filtering.

[Gemma Malagón]

Something like this?

function y = bandpass_filtfilt(Data)
%UNTITLED7 Summary of this function goes here
%   Detailed explanation goes here
%% Filter the signal using 4th order butterworth filter
n=size(Data,2);
[b a] = butter(4,[0.02 0.45],'bandpass');
for j=1:n
    x_filtered(:,j)=filtfilt(b,a,Data(:,j));
end
y=x_filtered;

[Star Strider]

Exactly!

The loop isn’t necessary, since filtfilt will filter the columns of your matrix by default. The documentation says:

filtfilt operates along the first array dimension of x with size greater than 1

There are other ways to design and implement filters. If you have R2018a or later, the bandpass function designs a very efficient ellliptic filter. It designs the filter and filters your signals at the same time. Use its second output (digital filter object) with filtfilt to filter all your signals. You only need to design it once if you want to use the digital filter object, if your sampling frequency and other requirements don’t change. It uses filtfilt in its filtering operation to produce the filtered first output.

[Gemma Malagón]

I tried using that code and is still not working, is there another way I could get rid of that flat line?

[Star Strider]

I am not certain what you want to do. The bandpass filter should eliminate the baseline.

If you are talking about the line at the right side of the plot that goes to zero and then back to (0,0), that is likely the way your vector is written. It likely looks something like this:

x = [0:10 10 0];
y = [rand(1,11) 0 0];

figure
plot(x, y)

so to get rid of the straight line at the lower end of your plot, you would have to get rid of the zeros at the ends of both vectors. I do not have your vectors to work with, so you need to experiment with that until you get it the way you want it.