Challenge 1: Make the eye tracker small enough to carry about, secure enough to prevent shifting with movement and discrete enough to not make a scene.

Desktop eye trackers include a monitor, a series of cameras trained on the eyes that either stand alone or are built into the monitor, and a recording unit of some kind, both for a steady stream of eye data and video capture of the monitor screen. Mobile trackers include the same, but now a monitor is not needed; instead, there is an additional camera unit, the scene camera. This camera records the scene as the participant encounters it and needs to be ‘attached’ to the user close to same plane and position as the eyes. This is potentially a lot of equipment that now needs to be carried around by the participant.

SR EyeLink w/ scene camera

One of the earlier mobile tracking systems that became available and was reasonably accurate was a modification of the desktop eye tracking system. SR Research took their EyeLink system and added a scene camera. This worked because EyeLink was not a remote system; it was a head-mounted unit and had room along the headband to host a scene camera. This head-mounted unit did not include the recording system and thus was tethered to a processor and hard drive. The cable was rather thick (thickness of a finger) and was limited to 40 feet. The recording system could be placed on a cart and with a long extension cord, could be pushed around after the participant. This system was certainly secure enough, but not designed for mobility or discretion. It was effective for small spaces, such as flipping through magazines, considering a display stand, or evaluating a single shelf set or package in hand. Nonetheless, the headgear was rather cumbersome and definitely drew attention.

Courtesy SMI

Luckily in the last 1-2 years there have been tremendous developments. Scene and eye cameras are significantly smaller and lighter and can be attached either to a cap or to a pair of glasses. Mobile trackers are still tethered; wireless systems are in the works, but so far the data streams are too heavy (with up to 100 data points per second and video from 2 cameras at 30 frames per second). Nonetheless, the cables are small, not much larger than those to your ear buds on your iPod. And, more importantly, they are tethered to equipment that is substantially less bulky - usually a recording device less than half the size of your typical laptop. This can be easily carried in a pouch that hangs over the shoulder of the participant or is otherwise attached.

Courtesy ASL

Which approach is more effective - glasses or wearing a cap? Glasses have certain appeal because they are smaller and less noticeable. With proper straps these glasses can be secured so vigorous head movement does not shift the cameras about. Camera movement can result in a significant and undesirable shift in the calibration (i.e. what the data or video indicates the user is looking at is no longer what the user is really looking at). Glasses are more easy to secure than a baseball cap.

But glasses pose certain problems. They cannot be used if the participant wears prescription glasses (believe me, we’ve tried!). Further, the positioning of the eye camera and cut of the glasses is designed for a certain face structure. Deviate from this standard and the edge of the opening cut into the lens falls between the camera and the eye, distorting the camera’s view of the pupil.

Wearing a baseball cap with cameras attached offers solutions for both of these

Courtesy SMI

challenges. The camera units attached to a cap are more flexible and offer more options for adjustment, allowing for accurate tracking of virtually any type of participant, young and old, with or without glasses, and any nationality. What about the camera shift? We’ve been reassured by the manufacturer that camera shift is monitored and seamlessly corrected via the tracking of the corneal reflection. If this is indeed the case, we’re sold! We have the opportunity to test out such a system in the coming weeks.

There are different mobile trackers available, and they differ not only in the hardware; some use dark pupil, some light, some with or without corneal reflection. Steps to calibrate, record and monitor in real-time varies by manufacturer. The robustness of the systems, especially if tracking in daylight or in particularly ‘bumpy’ environments (such as road car rallies!) varies. Detailed discussion of this will be dealt with in another post. For now, let me just say that not all mobile trackers are the same and do need to be carefully evaluated.