Eye tracking with a diffractive AR waveguide

Eye tracking is pivotal for human–computer interaction, yet current wearable devices have constrained detection capability due to side-view camera configurations. While cameras integrated with optical waveguides provide transparent front-view imaging windows, they typically necessitate customized structures to form clear eye images directly. In this work, we propose an eye-tracking system with a non-customized diffractive augmented reality waveguide utilizing a computational imaging approach. By employing computational reconstruction to address optical distortions, our scheme eliminates the strict requirement for hardware-based point-to-point optical mapping. The imaging quality is validated through quantitative metrics and downstream gaze estimation tasks, demonstrating an experimental gaze angular resolution of 0.776 ∘ under the present fixed-distance eyeball-model setup. This approach enables the reuse of commercial display waveguides for sensing, offering a compact and cost-effective path for advanced AR system integration.