whati's solution for signal filtering ?

Hi :) I'd like to know how to filter this signal in figure, can you give the solution ?

Thank's

[Jan]

It depends on what you consider as noise and what as signal. Maybe you want to get rid of the low frequencies, or the high ones. Maybe you want a band only, or exclude it. The readers cannot guess this detail.

[Mathieu NOE]

and also it's quite usefull to share the data and some code if you have started one

Answer 1

[Sarvesh Kale]

Hi Guendouz walid,

I have made the assumption that the data around time 100 and 350 is error data that you would like to get rid of, I suggest the following steps

1. subtract 400.2 from samples of data as noise data is having a mean of approximately 400.2, this will make noise data centered around 0
2. perform FIR filtering such that it will smoothen the data, FIR filter to be used can be 0.2*[1 1 1 1 1], FIR filtering can be done using simple convolution operation
3. Add the bias component back to the signal samples

Following is the code that achieves the above.

% assume data is X 
X = X - 400.2 ;
f = 0.2*[ 1 1 1 1 1] ;
filtered = conv(X,f) ;
filtered = filtered + 400.2 ;
plot(filtered) ;

OR

Alternately, there is also an option of Clean Data under Home tab of MATLAB.

Import your data and select the smooth function from the App.

[Jan]

Subtracting the estimated mean (why not calculating it?) reduces the transient effects at the margins only. If the initial and final signal differ, this is a weak approach.

With conv('same') a time shift is avoided. Using movmean instead offers a better handling of the margins, but the same values at the inside, without the need for assumptions about the DC level of the signal:

X = rand(1, 30) + 1000;
Y = conv(X, [0.2, 0.2, 0.2, 0.2, 0.2], 'same');
Z = movmean(X, 5);
plot(X, 'r'); hold('on');
plot(Y, 'b');
plot(Z, 'ok')
ylim([999, 1002])

But the main problem remains, that the OP did not clarify, what the noise and what the signal is. The blocks at X=100 and 350 look such similar, that they might be the searched signal.

[Sarvesh Kale]

Thank you for your inputs on mean calculation @Jan

[Guendouz walid]

thank you