Open dmurfet opened 2 years ago
The lack of even basic palm rejection was a problem in several sessions on 13-1-22, particularly in the morning Foundations and evening Cryptography sessions.
I don't actually have this problem under normal usage, but I can reproduce it by tapping the screen simultaneously with my left index finger and the Apple Pencil held in my right hand, to create the lines you see in the first image below
With the simple palm rejection added in ec915c262cf3f22b5a7a27ff990f99d03fa580bc most of these long lines are rejected (the rejection threshold is 100 pixels)
This isn't perfect, but observing our speakers in practice they usually generate lines longer than 100 pixels, because the distance from their pen tip to whatever part of their palm is triggering the touch event is further than that. I believe this simple fix will eliminate > 80% of occurrences. However, a smarter solution on top of this would be great.
I'm closing this for now, we can worry about improved palm rejection later.
This line is in the range we would expect to still be possible, but was generated "naturally" during an event. So we still need a better system for palm rejection.
I think we should pay attention not only to "palm lines" but also "palm dots" which are now a bigger problem.
Meeting suggestion: Halve the threshold
The multi-touch lines are gone now. All that's left is the dots drawn by your palm while your pen is lifted (dots while pen is down are gone).
Getting rid of those palm dots is less trivial - you need to track the historical positions of strokes and estimate the likelihood of a new touch appearing far away from them.
At the moment most users of the boards will have problems with "stray dots" or "stray lines" most likely generated by the lack of palm rejection. There may not be a perfect answer here, e.g. GoodNotes allows the user to configure how much palm rejection to do in software https://support.goodnotes.com/hc/en-us/articles/360000129755-Configuring-palm-rejection-in-GoodNotes-5.