Forums

Adaptive filter and noise reduction

Started by sh_ee 2 weeks ago9 replieslatest reply 2 weeks ago104 views

What limitation would I have if my maximum adaptive filter tap order is 8? how can I correlate this to the type of noises the adaptive filter can denoise?

Can the step size of the adaptive filter be adaptive itself, for example, increase when the filter is not converging to stabilize faster then be decreased again to "optimal" state?

[ - ]
Reply by JohnEhlersJuly 23, 2020

You may want to consider using an IIR filter.  Adaptive smoothing filters are used by technical analysis traders in Futures and Stocks.  You can Google such filter names as KAMA, VIDYA, and FRAMA.

[ - ]
Reply by sh_eeJuly 23, 2020

Wouldn't IIR be a lot more prone to instability than FIR? What benefit would it offer if I used it instead?

[ - ]
Reply by JohnEhlersJuly 23, 2020

Stability is not an issue if you keep the poles inside the unit circle.

[ - ]
Reply by dgshaw6July 23, 2020

Hi sh_ee,

I'm not sure that the information that you have provided is enough to answer the question.
Lets assume that you are talking audio, and that you sample rate is 48k, then the 8 tap adaptive filter will allow you to deal with noise correlation delays in the range +- 8 usec.
If we were looking at a two input system with a reference for the noise and an input for the signal to be de-noised, that would mean they are no more than 6 cm apart!

To answer your second question, you can always have a variable step size for the adaptation, as long as things remain stable.

[ - ]
Reply by sh_eeJuly 23, 2020

Hello dgshaw6, thank you for the answer! Sorry I didn't provide enough information. My sampling rate is 30kHz, my input signals are physiological signals and many types of noises can affect it (power-line, surge, power dips..)

I wanted to know how the filter taps limit the ability of the filter to denoise the different noises since they have different impulse responses.

Would you mind elaborating how you reached the +/- 8usec and the 6cm distance between ref and input?

[ - ]
Reply by dgshaw6July 23, 2020

I saw you additional question in another post, so I understand you problem a little better now.

I incorrectly assume that you would be working on audio, sampled at 48 ksps.  The leads to 8*1/48000 sec for the total span of the filter.
Assuming that you try to center the usage of the 8 taps, that give +- 8 usec correlation window.
Then, still asuming audio with 343 m/sec, that leads to 6 cm.

In the context of the 50 Hz noise you are concerned about for the other question, I have answered there.

[ - ]
Reply by sh_eeJuly 23, 2020

Great, thank you! that answers my question.

[ - ]
Reply by SlartibartfastJuly 23, 2020

Some easy generalizations suggest that any noise events longer than the filter may not be able to be cancelled by the filter.

The adaptation rate can certainly be adjusted on the fly, but be sure to consider how that affects the stability of the filter.

[ - ]
Reply by sh_eeJuly 23, 2020

Got it, so I am assuming wideband noises would be hard in my adaptive filter