Recommendation report for an infrared eye-tracking system for P-Cubee Leong, Ricky; Chan, Hei Wang
There are several techniques that can be employed to track real-time eye movements. Currently, pCubee uses magnetic sensor technology and requires the user to wear a head-tracking sensor that is wired to the device itself. The objective of the project was to construct a wireless, noninvasive, low-latency infrared eye tracking system that could replace the existing solution. The technique that was utilized in this project involved processing of images from a video-based method, which can be performed with inexpensive cameras. A two camera system was setup and calibrated successfully. Triangulation was tested with IR goggles and the accuracy of the system was determined to be less than 1%. Thus, proof of concept was demonstrated and it should not be difficult to extend the system to function with any two camera configuration, provided that the appropriate common field of view is produced. An algorithm was developed to search for the location of the red-eye effect in a subject’s eyes. However, based on the images gathered from the camera setup, the red-eye effect was not strong enough to be detectable. A difference image technique was also employed to improve the detection of the red-eye effect, but did not yield better results. There were many limitations or sources of error that prevented the system to be designed optimally. The investigators believe there are several improvements - such as utilizing higher quality cameras and better image processing techniques - that could be made to successfully complete the system.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International