Questions and Answers About the Owl


7/31/05

The Owl is an eye tracker developed by Martin King in the 1980's.  It measures relative eye gaze using a sensor ring that is typically mounted on a pair of glasses in front of one eye.  Some photos are on the Dasher page here:  http://www.inference.phy.cam.ac.uk/dasher/development/

(Dasher, by the way, is a very nice augmentative & alternative communication technique that deserves to be widely known.  Software that implements Dasher--open source, no less!--is available along with more information at:  http://www.inference.phy.cam.ac.uk/dasher/)

How does it compare to commercially available devices?

The first thing to realize about the Owl is that the sensor has optoelectronic elements distributed over a wide space.  The total area of silicon might be smaller than a CMOS imaging chip, but the packaging of many individual elements, combined with the dramatically different likely volume of production versus say a USB webcam, means that cost-wise, the Owl will be hard pressed to compete with image-based eye tracking.

In terms of performance, it's possible the Owl could compete with image-based eye trackers, but nothing is proven yet.

The Owl, like other head-mounted techniques, can only measure relative eye-gaze.  This happens to be an advantage for its primary application.

The bottom line:  The Owl is not the world's best eye tracker.  (See below for why anyone would want one.)

Can I Get One?

If you are looking for an eye-tracker, then the answer is that it is not ready for use.  Plus, there are other, better eye trackers out there.

If you are interested in a specific type of signal processing that you want to apply to the Owl data (e.g., neurally-networks), then I have a very limited number of prototypes of the hardware, and some data collection software that runs on Mac OS X.  (Another researcher has developed collection software that runs under Windows.)  Email me at dgrover at redcedar dot com.  Such work would need to be coordinated with other teams working with the Owl.

What is the current status of development?

I have several stable Owl prototypes built, and software that drives the Owl and collects data.  My analysis is done offline with the excellent Octave mathematical software.  (Octave is an open-source mathematics package that is similar, but in some cases nicer than, Matlab.)

Why develop yet-another-eyetracker?

In my opinion--and I am not the inventer of the Owl--the Owl may have a hard time competing with other eye trackers (primarily image-based) as an eye-tracker per se.  However, it has some characteristics that make it useful for a very specific application for people with communication problems.  In particular, it offers feedback to the user for a number of points of gaze, and if these correspond to keys on a reduced keyboard, the user can employ well-known techniques to generate, for example, English at the rate of one or more letters per fixation, with minimal cognitive load due to encoding or feedback.

When will it be available commercially?

Unknown if ever will be.  Certainly several years away, at best.  Again, though, it's not something I would look at as competing in any way with what can be done with cheap CMOS imaging.

Unless you're into AAC applications, the Owl is probably not what you're looking for.