Adaptive Deep Learning Approaches for Real-Time Gaze Tracking in Immersive Virtual Reality Environments
Amit Kumar Singh, Priya N. Mehta, Jian Chi, Leonardo G. Romero · 2024
Gaze tracking in virtual reality (VR) environments has become a crucial aspect of enhancing user experiences and developing innovative applications across various domains, such as gaming, education, and medical training. Despite the significant progress, real-time gaze tracking remains challenging due to the dynamic and complex nature of immersive VR environments. This paper addresses this critical research question by proposing adaptive deep learning approaches tailored for real-time gaze tracking within VR contexts. Our method leverages a combination of convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to process sequential eye-tracking data efficiently. We introduce a novel adaptive mechanism that adjusts the model based on user-specific behaviors and environmental changes. The primary contributions of our research include the development of a robust, real-time gaze tracking system that dynamically adapts to the user's gaze patterns and the VR environment's complexities. We further enhance accuracy by integrating attention mechanisms that prioritize relevant visual features. Experimental results demonstrate our system's superior performance in both speed and precision compared to existing benchmarks. Through comprehensive user studies involving diverse VR applications, we validate the system's adaptability and real-time capabilities. This research paves the way for more intuitive and responsive VR experiences, setting a new standard for gaze tracking technologies in immersive environments.