Obtain Residual value of last IMF

[nazmibojan]

Hallo @laszukdawid ,

I just open new issue because this is a different topic. I use EMD to implement denoising to an input signal. To reconstruct a new signal, we have to sum qualified IMFs and residual of EMD process. How can I get the residual value using this library? Thank you.

Regards, Nazmi

[laszukdawid]

Hey, If you take a look at the result of EMD you'll see that it returns a NumPy array. Simply select whichever IMFs you like. Quick example:

from PyEMD import EMD
s = np.random.random(100)  # generate random signal
emd = EMD()
imfs = emd(s, max_imf=3)  # Force max 3 IMFs
noise = imfs[0]  # First IMF
denoised_signal = np.sum(imfs[1:], axis=0)  # Sum of all but first IMF

[nazmibojan]

Okay, thank you for the answer, I think that code is very useful. But, the denoised signal are sum of qualified IMF + Signal Residual. As we know, the signal residual was the remainder/residual of last sifting process that produce last IMF. Or the last IMF is the residual of sifting process in this library? Thank you.

Regards, Nazmi

*PS: This paper describes denoising process using EMD.

[laszukdawid]

The last row in the array is the residue.

The noise is whatever you decide shouldn't be in the signal. Depends on the signal and model that generated it but typically it's the first IMF which is considered the noise. (Not necessarily the only one.) If you want to detrend then also remove residue (or other slow components).

[nazmibojan]

Great! Thank you very much @laszukdawid for this awesome library. Good luck for you.

Regards, Nazmi