EyeChat: An Eye-Gaze-Controlled Web-Browsing System for Head-Mounted Augmented Reality Displays

Accessing Web content using conventional input methods can be difficult for people with motor impairments. Current solutions often use webcams to enable eye gaze control, which may cause interaction delays and inefficient browsing. EyeChat is an eye gaze-controlled interface for efficient Web content delivery via Microsoft HoloLens 2. The system features a text entry interface optimized for augmented reality and a hybrid chatbot to deliver accurate content. By combining rule-based and large language models, EyeChat provides real-time, contextually relevant responses. Its frequency-based circular text-entry interface centralizes eye movement around head gaze to minimize effort. Predictive features, including current-word, next-word, and phrase suggestions, enhance typing speed. This paper presents an initial technical evaluation of EyeChat with 12 typically developed adults. Non-disabled participants reported that 83.83% of responses met expectations. The system achieved a text-entry rate of 9.88 words per minute and 0.53 keystrokes per character with minimal practice. Eye gaze analysis showed 95% of points fell within a 24° field of view.