I'm new to DSP and currently working on time-series data. The mentioned time series (of Toe) is extracted from a video tracking various body parts of an athlete. Ideally, there shouldn't be any movement in the signal and should stay around 0. I cannot explicitly make it zero, because, for other datasets, I'm not aware of what body parts are being tracked there; therefore, I need to do it in an implicit manner.

What are the noise removing techniques that can be used here? I used FFT transform, needless to mention it smoothened the signal.

enter image description here

  • 2
    $\begingroup$ Please add more details and make your question more specific. If the signal shouldn't have "any movement" and should stay around zero, why not just make it equal to zero? $\endgroup$
    – MBaz
    Oct 9 '21 at 16:08
  • 1
    $\begingroup$ How did you extract this data from the video and what are you trying to do with it? How did you use the FFT to smooth it out? $\endgroup$
    – Ryan
    Oct 9 '21 at 18:55
  • $\begingroup$ Although the extraction information is out of context for this video, my friend used Openpose. About FFT, I transformed the signal into the frequency domain, picked up frequency using the absolute value of the complex numbers. $\endgroup$ Oct 10 '21 at 14:50