2
$\begingroup$

I have just begun to dive into the field of signal processing, but there is the need to program a digital filter, that has to smooth a realtime signal from a sensor device. As far as I know, in my case a FIR filter is not suitable, because I have a lot of samples (1000) per sec and the convolution that is needed for appling a gaussian smoothing takes to long. Therefore I need some IIR Filter, that is capable of deliver results, that are comparable with gaussian filtering. I have already tested the first order lag filter, but this filter does not smooth out the signal if it was constant over time before.

Do you have any suggestions where to start? What are suitable fast recursive smoothing filters that look like gaussian smoothing? Some hints will be usefull for my to get the keywords for further reading.

Improve this question
$\endgroup$
2
$\begingroup$

Have a look at my Fast Gaussian Blur Project at GitHub.

You will find there implementation of IIR Approximation of Gaussian Blur which implements the following papaers:

  1. Recursive Gabor Filtering.
  2. Recursive Implementation of the Gaussian Filter.
  3. Boundary Conditions for Young - van Vliet Recursive Filtering.

The idea is pretty straight forward.

Improve this answer
$\endgroup$
  • $\begingroup$ Thanks, but I am not sure if the methods do work in my case. Im am not fully sure, but it seems that they need to know the signal in the first place, to go forward and backwards over the signal. Which in my case is not possible, bcause the signal has to be smoothed while it is read at the same time. $\endgroup$ – Lemonbonbon Feb 13 '19 at 11:56
  • $\begingroup$ What you can do is always treat what you have as the whole signal. Namely keep doing the first pass. At any new sample, start doing from it the 2nd pass to the other side. $\endgroup$ – Royi Feb 13 '19 at 12:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.