What digital filter(s) is this?

[GWestphal]

https://www.dropbox.com/s/3t254iidk306r59/Photo Feb 24, 16 56 12.png

Left: original signal. Right: Denoised.

I can see there is a high pass since below 1-2 hz is pretty attenuated.

But the alternating little attenuations that are very regular in time and frequency are perplexing me however.

What sort of processing causes that?

 

[Rstallard]

Do you have the raw data? Might try just looking at an FFT of the raw data. Is the heat map here blue - > red :: low -> high intensity?

 

[Anuextreme117]

Color Intensity, IDK

 

[uclabachelor]

You'll probably get more useful responses if you provide more information and context.

 

[herm0016]

spectrum analysis?

looks like an averaging filter of some sort.

like a triangle or gaussien over .5 sec or similar.

 

[KWiklund]

It could be an artifact of bad OLA sysnthesis? Can you provide more information about the signal, and how this filter is being applied?

 

[GWestphal]

Yes, those are spectrograms of the original and denoised signals.

These are ECG (heart beat) recordings, sampled at 200Hz, I know the original has should have some attenuation from near DC (0-1Hz) and higher (35+). But the additional processing from there is all digital.

 

[KWiklund]

I haven't done much with ECG signals, so I may be missing some nuances in terms of the signal structure. The striations seem to be in the original signal (non-denoised), so they don't seem to be a processing artifact from the de-noising processing. The other filters shouldn't cause that either.

I suspect it might be an artifact of the recording process, or possibly some interfering signal. It's clearly periodic and impulsive, but doesn't look "natural" to me. You might be able to learn more from the time-domain signal if you have it. Someone more familiar with this class of signals might also be able to tell you more about practical processing / pre-filtering issues.

 

[slugg]

Could you provide an FT of the original and filtered signals?