Waveguide-integrated eye tracking system

Abstract

We present a novel waveguide-integrated eye tracking system for smart glasses, offering a compact and aesthetically pleasing solution by embedding the eye tracking camera within the temple of the eyewear. The system features a front-view capturing capability, enabled by the waveguide, which minimize distortion and occlusion while ensuring consistent imaging conditions. Waveguide systems typically face challenges such as low light efficiency, which introduces white noise, degrading image quality and tracking performance. To address this, we employed lightweight image processing algorithms, including a neural network-based pupil segmentation model to extract elliptical parameters, center positions, shape, and orientation, from noisy, low-resolution images, enabling accurate and robust gaze estimation. In controlled benchtop experiments using an eyeball robot, the system demonstrated promising performance, achieving a mean gaze error of approximately 1 degree, validating its capability to maintain precision despite the challenges associated with waveguide optics. Variability in feature extraction was observed near the edge region of the target object space, emphasizing the need for refinements in waveguide design and feature extraction algorithms. Overall, the proposed waveguide-integrated eye tracking system represents a significant advancement by seamlessly incorporating eye tracking technology into everyday eyewear. Its design maintains user comfort and aesthetics while achieving the precision required for AR/VR applications, enabling intuitive and reliable user interaction.