Text Input Through Swipe Gestures Based Personalized AI Agent

Text input based on virtual reality is a common technique in interactive systems, but it may face challenges related to input efficiency and task load when interacting with VR hardware and applications. The development of artificial intelligence interaction technologies offers new approaches to addressing text input challenges. In this paper, we propose a swipe gesture-based text input method under a personalized AI agent framework, combining portable devices (e.g. smartphones) with VR input and incorporating user profile information, input habits, and conversational intent. By integrating the GPT-3.5 model to train a personalized AI agent, we emphasize the importance of understanding and responding to human behavior or capabilities from the agent's perspective, enabling text prediction based on specific contexts. The keyboard layout design is based on a disk divided into 8 equal regions, where the outer circle is subdivided into key areas containing letters, and the inner circle serves as the input buffer area. By resolving word ambiguities based on user input and leveraging the extensive capabilities of large language models in context awareness and text prediction, the system allows complete sentences to be generated from keywords. This reduces the number of manual inputs required by the user, improving text input efficiency and enhancing the overall user experience.