360 VIDEO VIEWING WITH VIRTUAL REALITY HEADSETS: CONNECTING USER HEAD MOVEMENTS TO INTENTIONS

In recent years, VR technology has significantly enhanced video immersion and efficiency. Numerous studies have delved into analyzing user panning, a head or body movement executed by an individual engaging with a 360-degree video, to improve video buffering by predicting such movements. However, this perspective tends to isolate user panning behavior as a distinct signal for modeling and prediction, neglecting the intrinsic value of the content and the user's engagement with it, thereby leaving a gap in comprehending the motives behind users' panning in these immersive settings.

Our research endeavors to bridge this gap by investigating the underlying reasons for viewers' panning behavior (head movements) in 360-degree content, employing thematic analysis of self-reported user intentions. This thesis illuminates the driving motivations for pans during 360-degree video interactions, as expressed by participants. Through categorizing the viewer's behavior with detailed annotations, we establish a structured framework for understanding viewer engagement.

Furthermore, we explore the feasibility of predicting viewer intentions based solely on video content, thus providing insights into content-driven viewer panning behavior. This work also introduces a meticulously compiled dataset that merges pan action data with corresponding user-reported motivations, presenting a valuable asset for subsequent VR viewer behavior research. Lastly, we unveil a user-friendly system compatible with any head-mounted device, designed to ease the replication of our study and enable real-time collection of user panning data, free from desktop tethering constraints. This system not only facilitates data gathering but also broadens the accessibility and utility of our research outcomes, significantly advancing the comprehension of human interaction within VR contexts.