Automated identification of areas of interest in dynamic head-mounted display videos for virtual reality forensic psychology applications

Högväg, J., Andersson, S., Nyman, T. J. ORCID: https://orcid.org/0000-0002-6409-2528 and Ventus, D. (2025) Automated identification of areas of interest in dynamic head-mounted display videos for virtual reality forensic psychology applications. In: HCI International 2025: 27th International Conference on Human-Computer Interaction, Gothenburg, Sweden, 22-27 June 2025. Lecture Notes in Computer Science. Springer, Cham. (In Press)

Abstract/Summary

This paper introduces a novel software solution to address the challenges of automated identification of Areas of Interest (AOIs) in dynamic, head mounted display virtual reality (VR) environments, with a focus on applications in forensic psychology. Traditional eye-tracking tools often require manual annotation of AOIs when analyzing moving objects, such as faces, in dynamic 360- degree VR scenarios—a process that is time-intensive. The presented software utilizes RetinaFace to generate consistent AOIs by dynamically tracking facial coordinates across video frames, accounting for variability in head movements. Outputs are seamlessly formatted for analysis in iMotions, enabling robust synchronization and visualization of gaze data. In a pilot study with 16 participants, validation against manual annotation confirmed the system’s high accuracy (92.2%). By automating a previously manual process, the software provides researchers with an efficient and scalable tool for analyzing complex visual attention data, significantly enhancing the feasibility of large-scale VR studies in forensic and psychological research. An empirical example illustrates how the software can be used.

Item Type	Book or Report Section
URI	https://reading-clone.eprints-hosting.org/id/eprint/121229
Refereed	Yes
Divisions	Life Sciences > School of Psychology and Clinical Language Sciences > Department of Psychology
Publisher	Springer
Download/View statistics	View download statistics for this item