I want to apply a Kalman Filter to the humidity data which I retrieve of the DHT11.

How can I do that?


Learn about the kalman filter and what it does. It requires a certain understanding of what it is, what it is good for and why you would eventually use it. It seems to me, that at the moment this isn't your strongest topic.

Thereafter, figure out which !multiple! sensor values you have, since only with a "model" and an "observation" a filter of this kind makes sense in the first place.

To specifically adress your question: you can apply any filter to any data. The hidden question, if it makes sense in this case is not trivial to answer, since you didn't exactly provide the vast information. My guess is, you only have this one sensor, and it would be (at this state of knowledge) a reasonable idea to use a bandpass filter of sorts instead.

| improve this answer | |
  • Hi, thanks for your reply. I have some knowledge about Kalman filter in theory. But i didnt yet apply anwhere and I have some data in my figures. The humidty is acting +-%5 and I want filter this data. As i see on internet, some people use Kalman filter to temperature and humidity data for best results. I can research about applications of Kalman filter. May be someone help me for earn time. – Adem Gül Mar 28 '16 at 20:57
  • As I said before, you didn't give enough information for anything. Please state a specific question. If you know, what a kalman filter is, you need to think about what your observation is, and what your model is. Then you need to state the covariances of both of these and then you implement the filter. However, it seems to me, a lowpass is a better fit for a one-sensor problem. – mike Mar 28 '16 at 21:04
  • 1
    Thanks, may be I should basicaly implement a smoothing filter for DHT data. – Adem Gül Mar 28 '16 at 21:38
  • That's what I suggested. I think a kalman filter is a very nice tool, but it has it's place and it's nowadays used, where it really has no place at all. Just because it became famous, doesn't mean it does wonders for each problem. But still, all of this is guessing, since you didn't specify, what resources you have, or what your goal is. If you have many sensors and a lot of time, a kalman filter might just be the thing! – mike Mar 28 '16 at 21:46
  • 2
    Btw: If you calculate a discrete time kalman filter for a linear system with only one measurement, the formulae automatically yield a simple lowpass filter (the steady state KF). This means, in the sense of the least quadratic error, it is best to just use a lowpass. – mike Mar 29 '16 at 6:55

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.