Two new SeeingMachines eye trackers have arrived to the P54 lab. They have successfully survived their long trip from Australia in their well rugged cases.
We have already one eye tracker from the same company installed in our driving simulator. This tracker has provided us with lots of interesting new data on driver behavior. Some of the results of the eye tracker usage can be found in my Master’s Thesis. We have also submitted two papers for conferences with data from the eye tracker. We also have a number of prototype experiments designed using the tracker.
With the two new systems we hope to
1) increase the quality of tracking in the simulator
2) have one system installed in a real car for naturalistic driving experiments
3) use one system part-time on different other devices for user interface research
The quality of eye tracking in the simulator can be increased with additional systems that can be connected using the faceLAB Link application. Installing cameras near a GPS screen or near the central console will give us more precise tracking information on these locations.
Two eye trackers’ output can be integrated into a single stream of data using faceLAB Link, which is a part of the faceLAB package. The computers of these eye trackers must be connected via network, and faceLAB must be running on both of them. After that, faceLab Link can be started on one of them. It automatically recognizes all trackers on the network. There is no need for separate calibration, because the application calculates the location of the two eye trackers automatically, when they are tracking the same face. When the data is fused together, it can be found in Link’s logging folder instead of the faceLAB logging folder. In order to have correct object intersection data, the data streams must be ‘re-coded’ using WorldView (another application in the package). This must be done, because the ‘slave’ eye tracker cannot produce gaze intersections with the virtual world objects of the ‘master’ tracker. But loading Link’s logged data and the world model of the ‘master’ system, and then re-coding it by pressing on the ‘record’ button, produces object intersections of the integrated data stream with the master’s world model. The output of this re-coding is recorded in WorldView’s logging folder.
We are very excited to use these systems in our research. The results of these experiments will be published on this site, so check back soon.
Oszkar Palinko