| Objective: ||Develop a set of general algorithms to correct for systematic error of different kinds in eye point-of-regard (POR) data, and a software tool to support applying those algorithms to diverse datasets.
|| Description: ||Data about where an individual is looking can provide valuable information about cognitive processes, attention, and strategy (Gluck, 1999). The resolution of eye point-of-regard (POR) data is much more dense (usually c. 60Hz) than inputs that are made to computer-based task environments. Key presses and mouse movements may occur sparsely across seconds or minutes as an individual completes a task. By including data about where the eye is looking on a moment-to-moment basis we can fill in details about cognitive process, which can serve as the basis for instructive interventions and evaluation.
Eye tracking technology has existed for decades, but the sophistication of these systems has increased dramatically in recent years. POR estimates can be made without restricting head movement and without requiring cumbersome equipment. Modern eye tracking hardware can easily be attached to the brim of a baseball cap, and is lightweight enough so that it does not cause discomfort or neck strain. To make optimal use of this source of data, however, requires that the POR estimates be accurately associated with regions of interest on the screen. All eye POR data has biases and error, both as a function of the geometry of the eye and limitations in the computational algorithms that are available for estimating POR. One technique for compensating for systematic error in POR estimates is to apply a correction algorithm to the data, adjusting POR estimates based on regions of interest in the visual field (Douglass, unpublished; Hornof & Halverson, 2002). While this technique shows promise, no general approach exists for applying a corrective algorithm to eye POR data, nor has a careful evaluation of alternative algorithms been performed. This proposal is for developing a general approach to correcting for systematic error in eye POR data.
|| ||PHASE I: Develop a software implementation to conduct error correction on POR data in different forms, and provide for using different algorithms to conduct the correction.
|| || ||PHASE II: Validate the different algorithms for performing POR error correction and establish criteria for using different options. Develop a comprehensive implementation that handles multiple data formats (including user-defined formats), implements a number of alternative correction algorithms, and includes a decision aid for selecting the appropriate technique for a given protocol/dataset.
|| ||DUAL USE COMMERCIALIZATION: Military application: This technology will increase the pace of progress in analyzing & interpreting human eye data in military research on human cognitive processes, in applications from mission planning to piloting. Commercial application: This technology can be incorporated into eye tracking systems to provide real-time error correction capabilities. Eye data is commonly utilized in domains including marketing and academic research.
|| References: ||1. Gluck, K. A. (1999). Eye movements and algebra tutoring. Doctoral dissertation, Carnegie Mellon University, Pittsburgh.
2. Hornof, A. J., & Halverson, T. (2002). Cleaning up systematic error in eye tracking data by using required fixation locations. Behavior Research Methods, Instruments, and Computers, 34(4), 592-604.
|Keywords: || Eye tracking, Point of regard, Experimentation|
Error correction, Bias, Algorithm