Open Input Framework Gesture Algorithm and Definition Tuning for Natty

Registered by Duncan McGreggor

By UDS, people will have had a chance to work with our gesture engine. What's going well? What should we do more of? What's broken? What needs fine-tuning?

In particular, we want to focus on these two areas:
 * how well are gestures being recognized and parsed by the engine?
 * how are users finding the actual physical gestures defined for Maverick?

Note that some issues will be hardware/firmware problems, kernel driver problems; how can we mitigate these? how can we get folks clued into tools such as the N-trig calibrator for Linux that Rafi wrote?

One example, from Henrik:
"...the two-finger pinch. At first, we wanted to use a focus point for propagation, such that a point in between was were the events would go. Then we learned that it might be better to use the common visible branch of windows, which often means coming down to the root window. But I realize that when placing one finger first, in a window, and putting a second down to zoom, it does not matter where that [second] finger is."

Tags: hci-n, hci, touch, multi-touch, multitouch, gestures, oif stack, oif, algorithms. optimization

Blueprint information

Duncan McGreggor
Chase Douglas
Series goal:
Accepted for natty
Good progress
Milestone target:
milestone icon ubuntu-11.04-beta-1
Started by
Duncan McGreggor

Related branches



Henrik has been engaged in this already, tweaking grail as needed. We haven't received any bug reports about gesture interpretation yet, but we're keeping our eyes peeled for anything. Grail's getting a bit of a makeover as part of the Natty architectural changes. I believe Henrik has identified some tuning work he'd like to do in that process.

Work items:
[rydberg] specific tasks for tuning in Natty: DONE
[rydberg] switch from a tap primitive to a touch primitive presents an opportunity to further reduce the number of parameters: INPROGRESS
[rydberg] Look into glue times per touch-up/down and number of fingers, to overcome typical false positives as a multi-finger gesture ends: TODO


Work Items

This blueprint contains Public information 
Everyone can see this information.