Gaze Pattern Genius: Gaze-Driven VR Interaction Using Unsupervised Domain Adaption
Kexin Wang, Yang Gao, Wenfeng Song, Yuecheng Li, Aimin Hao · 2024
This research advocates shifting VR interaction to gaze-driven interaction, a more intuitive alternative to traditional controls like VR controllers or gestures. Our focus is on enhancing neural network recognition accuracy, especially with limited user-specific gaze data. We introduce a novel framework for capturing gaze gesture patterns and propose a template dataset concept to boost neural training. Our unsupervised domain adaptation model, blending template depth and sparse user data authenticity, consistently excels in recognizing gaze patterns across diverse users. Rigorous benchmarking against leading architectures consistently shows our method outperforming. Empirical user studies confirm: gaze-driven interactions not only elevate VR experiences but also redefine immersive VR control dynamics.