Improved Accuracy Test Method for Mobile Eye Tracking in Usability Scenarios

Abstract

Eye tracking has been used in usability testing for many years to gain objective measurements to inform label, instruction, and product design. With many different testing environments, possible participants, hardware, and software, key metrics can vary greatly. These metrics can also vary for different studies, so a standardized test method, metrics, and calculation method are proposed in this study. The Tobii Pro Glasses 2 is a mobile eye tracker that does not significantly affect user mobility compared to other eye-trackers. This study aims to build a testing method, which can modify to better fit the varying conditions found in usability testing. This study was performed with Tobii Pro Glasses 2; however, this test method can be used with any mobile eye tracking units. Even with poor testing conditions, this test method results in reliable metrics which can be utilized to inform expectation and decisions with eye tracking. These methods are recommended to be performed prior to eye tracking testing to determine testing-specific performance.

Development and validation of a new methodological platform to measure behavioral, cognitive, and physiological responses to food interventions in real time

“…Low gaze capture may result from incorrect calibration vs. actual distance to the stimuli (MacInnes et al, 2018 ), incorrect positioning of the eye-tracking glasses (i.e., resulting in the front-view camera being directed away from the AOIs) or inherent inaccuracy (e.g. participants with deficient visual acuity, uncontrolled head movement) (Thibeault et al, 2019 ). We screened participants for sufficient visual acuity and performed individual calibration for each subject in the actual testing room, following the manufacturer’s instructions which involve a one-point calibration method with a card placed at comparable distance to the actual stimuli.…”

“…In our study, 73% of recordings provided 80% or more of gaze capture. Based on the sample size ( n = 149 recordings) we believe this limitation is acceptable, however, pre-test point-to-point and trial-to-trial variability checks would probably improve accuracy and are recommended (Thibeault et al, 2019 ).…”

“…Despite this, the coding necessitated frequent AOI adjustments due to the combination of head movements and the AOIs changing in size and shape over time. To improve accuracy, future tests should include a pre-test point-to-point and trial-to-trial variability analysis (Thibeault et al, 2019 ). Categorizing subjects according to natural head movement pattern while eating may also assist in feature matching AGM analysis.…”

Evaluating the integration of eye-tracking and motion capture technologies: Quantifying the accuracy and precision of gaze measures

“…Aside from establishing the origin of the gaze vector (e.g., eye position), expressing point of gaze with a world-based frame of reference can be affected by the accuracy and precision of the eye tracking system ( Ooms, 2015 ). Multiple factors influence gaze calibration and data quality, including task factors such as viewing distance defined as the distance between a participant and the location being fixated (such as Figure 1 ‘C’) and viewing angle defined as the angle between gaze origin and the location being fixated (such as Figure 1 ‘ a ’) ( Blignaut & Wium, 2014 ; Thibeault et al, 2019 ). Parallax error represents a common limitation associated with viewing distance and head-mounted eye-trackers, which occurs when the distance between the point of regard and the user (viewing distance) is different to when the system was calibrated.…”

“…Although the combined influence of viewing distance and visual angle outlined above are simple to translate into world-space values using trigonometry, the accuracy of eye-tracking systems (in terms of visual angle) has also been shown to be influenced by a range of participant factors such as eye color, contact lens use, and eye lash length (e.g., Blignaut & Wium, 2014 ; Thibeault, et al, 2019 ). Furthermore, MacInnes et al (2018) also reported that accuracy decreased as viewing distance increased.…”