Gunnar Beutner

Hm yeah, this patch doesn't really solve the problem for non-ESPresense devices anyway.

Also, this patch is entirely obsolete now that I've figured out a way to get the channel for received advertisements:

Feel free to give this a shot: https://github.com/ESPresense/ESPresense/tree/channel-info There are still quite a few TODOs though. Right now it doesn't handle devices which only advertise on one or two channels...

I'm not _quite_ happy with this yet - will update the PR when it's done. :)

I've had another look at the old code and I think there are three problems: * The bug in the one_euro_filter class (i.e. `hatxprev` vs. `xprev`). * I've just realized...

Shouldn't those constructor params be the other way around? ```c++ FilteredDistance(25, 0.1) ``` With the params the way they were the filter output is virtually identical to the input data:...

Oh yeah, I noticed that as well. Try this: ```patch // Update covariance - covariance[0][0] += deltaTime * (covariance[1][0] + covariance[0][1]) + processNoise; - covariance[0][1] += deltaTime * covariance[1][1]; -...

I've spent the past couple of days reading research papers after denoising algorithms. I ended up half-way through an implementation of a denoising auto-encoder when I think I realized something....

Also, I wonder how well the RSSI dispersion pattern is correlated with specific points in space. Maybe we can use that fingerprint to distinguish the angle of the transmitter towards...

Awesome, thank you! About the variance: That sounds useful - I'll add that to my filter and play around a bit with it a bit. This might be even more...