Paul Jolly^1,2, Anis Idrizović^1,2, Soma Mizobuchi^1,2, Ruei-Jr Wu^1,2, Michele Rucci^1,2; ¹University of Rochester, ²Center for Visual Science, Rochester, USA

Measuring eye movements with high precision is crucial in vision science. Building upon the classical dual Purkinje imaging (DPI) method devised by Cornsweet and Crane in 1973, we have recently developed a digital DPI eye-tracker (dDPI; Wu et al, Journal of Vision 2023). This system leverages modern digital imaging technology and computational power to yield robust measurements with sub-arcminute precision while circumventing several limitations of the analog device. To achieve high-resolution, the dDPI requires immobilization of the observer’s head, which limits the range of possible applications. Here we present a head-mounted dDPI, a system that enables precise oculomotor tracking while leaving the head free to move normally. The system uses a high-speed camera to image the first and fourth Purkinje reflections (the reflections from the cornea and lens) of an infrared beam. A GPU-based algorithm estimates the relative positions of the Purkinje images. The apparatus is capable of resolving arcminute rotations, as demonstrated by means of both artificial eyes and measurements from real eye movements. When coupled with a motion capture system that tracks head movements, this device yields an error in localizing the line of sight during normal head movements smaller than a few arcminutes. In sum, a head-mounted dDPI appears to be a valuable tool for studying active vision under natural conditions.

Acknowledgements: This work was supported by National Institutes of Health grants EY018363 and P30 EY001319.