Abstract: We introduce a proportional-integral-derivative (PID) controller based method for optimizing the calibration process of eye movement tracking systems. Our approach reduces the deviation in the correlation between the user pupil and the eye tracking device output. We apply our PID Error correction scheme continuously (i.e. continue on calibration also after an initial calibration procedure is terminated), this reduces the time (and the actions) required for calibration to minimum.
In eye tracking control systems the collected information can be used for a variety of fields such as marketing [1], technology solutions [2], and improved weapons systems in the military market [3]. The eye tracking data provides information on the subject's attention which can be used for studying cognitive processes [4]. Still, due to eye structure differences it is essential to calibrate the system prior to using such systems. Most of the eye tracking calibration procedure shares a similar idea: A camera detects the pupil; the user is exposed to a discrete set of points on the computer screen which the user is asked to follow. Obviously, a misaligned, lost of user focus or poorly calibrated system can produce wildly erroneous data.
Thus, we propose a calibration method which performs a continuous correction using a PID controller loop. The controller considers the error value as the difference between the measured pupil fixation position (corrected according to prior steps) and the algorithm's indicated position (i.e. the loci of the screened points) and makes a correction by adjusting the calculated error (vector) with the the PID constant values. The calibration procedure is performed as follows: The user is asked to follow a fixed smooth pattern projected on the screen (an eight figure and two concentric circles). The algorithm measures the differences between the output data in the patterns and calculates the PID values.We examine two open loop methods:Ziegler-Nichols Method andthe Cohen-Coon method for calculating the PID parameters Kp, Ti and Td.Once the PID parameters are obtained, one can continuously correct the error.
Experiments indicate that the aforementioned method is fast. Moreover, patterns which are somewhat more natural to follow reduce the chances of losing focus during the calibration process making the calibration results more accurate.
REFERENCES
[1]P.J. Chandon, W. Hutchinson, E.T. Bradlow and S. Young. (2007). "Measuring the value of point-of-purchase marketing with commercial eye-tracking data",Wedel, M. & Pieters, R. eds. Visual marketing: from attention to action(pp. 225-258), Mahwah, NJ: Lawrence Erlbaum Associates.
[2]A. Poole and L. J Ball. (2007). "Eye Tracking in Human- Computer Interaction and Usability Research: Current Status and Future Prospects",Chapter in C. Ghaoui (Ed.): Encyclopedia of Human-Computer Interaction.Pennsylvania: Idea Group, Inc.
[3]T. Schnell. (2000). "Applying eye tracking as an alternative approach for activation of controls and functions in aircraft",A Technical report, Iowa Space Grant Consortium's.
[4]M.Hayhoe and D. Ballard. (2005)."Eye movements in natural behavior",Trends in Cognitive Sciences, University of Rochester. Vol.9 (No.4)
