Very difficult waveform, having trouble smoothing it

Hi all, I posted a question last week related to this, but after several days of trying different things I still made no progress.

Essentially I am interested in eliminating the mini-peaks in this dataset provided, where after filtering, what is left is a a smoothed-out, sinusoidally shaped wave. I have tried things such as sgolayfilt, medfilt, curve fitting, but still have no had any success.

Any help would be greatly appreciated, thanks!

[Star Strider]

I don’t have a solution but I do have a suggestion. Plot a spectrogram of it to give you some insight into the problem. You data have a ‘chirp’ quality that will make it extremely difficult for you to determine what is signal and what is noise.

figure(1)
spectrogram(temp_data1)

Answer 1

[ag123]

I also tried taking the fft and multiplying it with a band-pass filter square wave, but this didn't seem to help much.

Answer 2

[Image Analyst]

I don't think any linear filter, or standard non-linear filters, will work on your signal. They generally operate on the whole signal and do the same thing everywhere. Since you have a very specific shape superimposed on only some portions of the signal I think you will have to write some very specific ad hoc code to recognize and "fix" those segments with the "bad" signal shape, while leaving the rest of the signal untouched.