Page 1 of 1

accuracy problem

PostPosted: 21 Sep 2014, 12:24
by edolgunsoz
I am having serious accuracy problems despite perfect calibration. The mouse cursor is not completely accurate and slowly going different sides. Also it is flickering too much. What do you think?

Re: accuracy problem

PostPosted: 21 Sep 2014, 23:28
by PixelHero
With a 5 stars calibration, I have very high precision, the mouse cursor is constrained to an area maybe 2-3 mm across and (when smoothed) low travel in that area, but extremely low accuracy. The cursor might be an inch or more away from where I am looking. Not drifting, just sitting there.

Ultimarly, that's a good sign. Low accuracy can be fixed by math, low precision can't.

They have a mythical 1.0 build in QA that should improve accuracy, but in the meantime there isn't much you can do, other than roll your own calibration technique and cursor control.


With a lot of care, you can get pretty close, but as of right now, the accuracy is just not high enough for direct computer control.

Re: accuracy problem

PostPosted: 22 Sep 2014, 00:21
by JeffKang
the accuracy is just not high enough for direct computer control


Even with the accuracy now, there’s some potential.

Eye tracking two-step-selection process + keyboard: there is potential for eye tracking features that would allow an eye controlled cursor to snap, zoom, etc. to a smaller target element, or make smaller elements project into large elements.

(My possibly stupid idea: Tag nearby elements w/ color & ID (like the Vimium Chrome extension) – project to large elements: http://i.imgur.com/3erfG6K.png)

Eye tracking teleport + mouse and keyboard: However, whenever you need the mouse, eye-tracking will still be there to provide an initial cursor teleport.
Eye tracking can be used to initially teleport a mouse-controlled cursor near an intended target.
Once there, the mouse can override eye-control when precision is needed.
Eye gaze is for initial, instant, and possibly large cursor movements.

A paper called “Mouse and Keyboard Cursor Warping to Accelerate and Reduce the Effort of Routine HCI Input Tasks” evaluates how initially teleporting the cursor with eye tracking in other common human computer interaction can affect the interaction.
A competition pits a mouse vs. an eye tracker + mouse.
You can see the performance of the eye-tracking warping + mouse at 2:41 of the video: http://youtu.be/7BhqRsIlROA?t=2m41s.
“Mouse control + eye-tracking teleport” ends up being the clear winner.