Support for multiple EyeX Controllers

[Sarah Lee]

We want to use at least two EyeX controllers on the same machine.
It would be most useful if the EyeX API would allow the developer to specify whether multiple trackers are:
* synchronized “in-phase” (i.e., sampling at the same time) to maximize accuracy (by averaging)
* synchronized “anti-phase” (i.e., optimally spacing samples) to maximize the effective sampling rate
* unsynchronized

[Anders]

Hi Sarah, thanks for your input!

How would you mount the trackers — one on each display, or in some other configuration?

[Peter]

I’m thinking the idea would be to put one at the bottom of the screen and one at the top to collect more data, giving more accurate angles or faster sampling rate. The samplig rate is probably high enough for most cases.

I do think there would be quite a large crowd that would be interested in being able to cut down on that roughly 1 or less than 1cm tracking area on the screen.
It would get it closer to mouse level accuracy for interaction without the need to zoom. Assuming it works.
When different companies start releasing their products a lot of people will be looking for stats and reviews on accuracy of the setup.

I think SMI had their RED-m or RED-n something on the way.
http://www.smivision.com/en/gaze-and-eye-tracking-systems/products/redm.html
http://www.smivision.com/oem-eye-tracking/products/red-n/index.html
It looks very similar, is it the same hardware? Sounds like they are releasing it this summer.

Anyway, I have a feeling that the deciding factor for consumers picking products will be accuracy, reliability ease of use.

Extra features that would allow increased accuracy without zooming in first could be the deciding factor on the consumer market. Anyway, it’s an interesting device that could help free up your hands for other input devices in the future.

[Robert [Tobii]]

Hi Peter,

One of the big challenges for bringing eye tracking technology to the consumer electronics market is to maintain reasonable performance while keeping the production cost low. The idea with double trackers in sync to increase accuracy or sample rate is nice, but it will cost twice as much for the end user.

Thank you for the links. SMI is developing their own hardware. It will be exciting to see how their RED-n performs, and how much it costs.

[varivera]

Dear Anders,

What if I want to mount two eye trackers one on each display? How would the calibration process be?
My idea is to connect two eyeX to the same machine. Since the targets are different (different displays), the result of each calibration will not be affected. Will the current calibration process, using the Tobii EyeX SDK, allow me to specify which EyeX I am calibrating? should I use Tobii Gaze SDK?

PS: where can I find more information about this?

Thank you

[Anders]

Hi varivera,
currently both the EyeX Engine and the EyeX Controller are limited to one tracker per machine. But the Gaze SDK doesn’t have this limitation, and the REX eye tracker doesn’t either. So if your application demands it, a Gaze SDK-with-double-REXes setup is what I would recommend. (Or one EyeX Controller and one REX.) This is clearly not as easy to work with as the EyeX SDK but it can be done.

Please check out the Gaze SDK developer’s guide, particularly the parts about display setup and calibration, for more information. It’s a pdf document included in the SDK package which can be found on the downloads page.

You will probably see some interference between the eye trackers leading to a percentage of the gaze data packets being lost. We have tried this and found the data loss to be acceptable. (Sorry, I don’t have any numbers I can share. It was too long ago.)

[Author]

Posts