Experimental procedure for metrological characterization of AR-based eye-tracking interfaces

Given the increasing demand for hands-free input interfaces within Augmented Reality (AR) applications, this paper addresses an experimental characterization of eye-tracking technology as an input mechanism for AR Head-Mounted Displays (HMDs). To this end, an AR application was developed with the aim of simulating a real-world application scenario. In this scenario, a set of objects is rendered within the scene so that each object can associated with a distinct command to be executed. The purpose of the developed application is to assess the capability of the input interface in accurately recognizing the objects selected through ocular movements. This evaluation also encompasses the interface performance in detecting the user-declared point of gaze, thereby quantifying the error between the user-reported focus and the interface perceptual outcomes. As a case study, without loss of generalization, the AR HMD Microsoft HoloLens 2 is considered. Eight different subjects were involved in the experimental campaign. The obtained experimental results showcase satisfactory performance with HoloLens 2. This paves the way for more robust development of eye-tracking-based applications, even in scenarios with stringent requirements.