Human Attention in Virtual Reality Applications: Exploration of VR Artwork

Abstract

Virtual Reality (VR) technology is revolutionising the way we interact with digital content, offering immersive and interactive experiences. Despite its advancements, challenges persist in optimising user experiences, particularly in the domains of attention and locomotion. This thesis aims to explore methodologies and techniques for investigating users' attentional focus and locomotion patterns within virtual environments, presenting novel approaches to enhance VR user experiences.
The research begins with an exploratory study, investigating the area of VR, hardware technologies, tracking techniques, and identifying research gaps. Skill development in gaming, data science, machine learning, and virtual reality prepares the foundation for the subsequent research phases.
A comprehensive research design is employed, utilising a custom-built VR user experiment which captures the eye gaze and body movement patterns of 35 participants interacting with a purpose-built large-scale abstract VR painting. This rich dataset facilitates the analysis of users' attentional focus and locomotion patterns.
To achieve the research objectives, innovative approaches are introduced. The VR heat map, consisting of opacity-based and saturation-based visualisation techniques, effectively visualise users' attentional allocation within the virtual environment. These methods capture the nuanced aspects of attention and locomotion, enabling comprehensive assessment.
The study reveals a strong correlation between users' behaviour in VR and their background characteristics. Notably, gender differences are observed in head and eye movements, prompting the development of a gender classification model for customising services and investigating potential biases in VR-related technologies and application design.
Deep learning models are employed to effectively model and predict users' locomotion patterns. LSTM-based deep learning models capture the complex patterns of user locomotion in VR, offering insights into user navigation and free walk exploration. By integrating recommendation navigation tools guided by locomotion modelling, users' navigation is enhanced, resulting in increased walking distance and coverage area within the virtual environment.
The impact of guidance techniques on users' viewpoints is investigated. Recommendation tools provide personalised guidance, influencing users' attentional focus and overall VR experience. Participants demonstrate improved exploration and engagement, indicating the effectiveness of the guidance tools.
This thesis contributes to the field of VR immersive by addressing research problems, providing insights into data acquisition for studying attention and locomotion, exploring the relationship between user behaviour and background, and developing modelling techniques to enhance user experiences in VR. The findings pave the way for future advancements in understanding and improving user experiences within virtual environments. The results can benefit the future of VR development applications.

Date of Award	27 Mar 2024
Original language	English
Supervisor	Mu Mu (Director of Studies), Suraj Ajit (Supervisor), Gary Hill (Supervisor) & Tawfiq Al-Assadi (Supervisor)