how to filter the noise of a signal

Hi, I need to filter the signals when the value is close to 0 (where the noise is much present). as you can see from the picture there is a lot of noise. I attached the .mat file and the picture

I don't need a particular filter, i just need to clean it a bit so I ask you which is the best method.

thanks.

Answer 1

[Mathieu NOE]

hello

I was surprised to see that your signal length is very long (730185 samples) and probably you are using a quite high sampling rate for what you really need

so basically I started downsampling your data with decimate (that includes also a low pass filter) and that is enough to already decrease significantly your noise level (which is high freq noise)

as a second step you can add some smoothing with smoothdata (equivalent to a low pass filter without phase distortion)

fyi, you can remove the decimation step if you really need to keep that high amount of data, but for plotting we rarely need more than a few thousands points

Zoom

load('matlab.mat');

% convert to double
F = double(F);
x = 1:numel(F);

% decimate first
r = 25;
Fd = decimate(F,r);
xd = 1:r:x(end);

% then low pass filter 
N = 25;
Fs1=smoothdata(Fd,'movmedian',N);

plot(x,F,xd,Fd,xd,Fs1)
legend('raw','decimated','decimated and smoothed');

Answer 2

[John D'Errico]

This is probably a good task for a median filter.

load matlab.mat
whos

  Name             Size              Bytes  Class     Attributes

  F           730185x1             2920740  single              
  ans              1x35                 70  char                
  cmdout           1x33                 66  char                

plot(F)

Fhat = medfilt1(F,100);
plot(Fhat)

Larger values for the second argument will produce a wider window for the filter.

There are many options of course, and many tools you can use.

Answer 3

[Joel]

toleranz=5;

for i=1:length(F)

if abs(F(i))<toleranz

F(i)=0;

end

end

plot(F)