0

I am trying to learn about signal processing and fft's applications within it. My understanding is that a classical use case for fft is to remove "noise" (ex. high frequencies that are not supposed to be in the original). However, while on StackOverflow, I came across this post that seems to suggest that this is a bad idea.

Am I misunderstanding the post or fft's use cases? If so, how would it be used?

vvm32812
  • 1
  • 1
  • click on the [tag:fft] you've used – you'll find hundreds of questions revolving around applications of the FFT; you're question is really just asking for a practically infinite list, and thus is a bit broad.. – Marcus Müller Feb 29 '20 at 21:00
  • Welcome to SE.SP! For the specific question you link to, there is already a question on this site as to why that particular method for filtering using the FFT is a bad idea. That is why I've closed the question. Even more, the FFT will not remove noise in a sensible way, except in very specific circumstances. There are other methods for noise reduction that work better and more generally. So I can't see how this question can be answered as you've asked it. Feel free to edit it and make it more specific, but as it's not answerable. – Peter K. Feb 29 '20 at 21:23
  • @PeterK. What are other "classical" uses for it? I generally prefer learning through use cases rather than the many textboooks linked throughout the site. – vvm32812 Mar 01 '20 at 01:41
  • @MarcusMüller Most of the stuff posted here assumes some base understanding of what it can be used for and is more asking about implementation to the best of my understanding. Where would I be able to find "classic/textbook" use cases of fft as it relates to signal processing? – vvm32812 Mar 01 '20 at 01:43
  • basically, all introductions to signal processing at some point extensively deal with the DFT (and thus, the FFT). Just start somewhere. Maybe MIT OCW on signal processing helps: https://ocw.mit.edu/resources/res-6-008-digital-signal-processing-spring-2011/ – Marcus Müller Mar 01 '20 at 09:28
  • The FFT is deterministic as opposed to stochastic (random). Most noise is stochastic. To deal with noise appropriately, you generally need a stochastic algorithm. The FFT is a fast way to decompose a time domain signal into its component frequencies. As such, its prime purpose is for filtering: altering the gain of the different frequency components. Now, if your noise is frequency selective, you can use the FFT to deal with it by filtering. But that’s not the prime use case for the FFT because that’s usually not why you need to filter. – Peter K. Mar 01 '20 at 13:13

0 Answers0