Dataset for Seated Body Leaning Pose Estimation from VR Headset Data

This dataset contains sensor and annotation data from 14 participants in a study of body leaning behaviors during immersive virtual reality (VR) job interview simulations. Each participant (anonymized as a numeric ID) completed two sessions with the VR Interview application, which differed in the timing and magnitude of the virtual interviewers’ head nods and other nonverbal behaviors. These two sessions are labeled <i>"Session 1"</i> and <i>"Session 2"</i> in the dataset folders for each participant's data. Each session folder contains time-stamped annotations of leaning events and synchronized time point-by-time point sensor data capturing head and hand positions, orientations, and gaze vectors.The <i>VR-Data.csv</i> files provide detailed time-series data for each session, including 3D positions and orientations of the headset and controllers, as well as gaze information. The accompanying <i>Timestamps-Leans.txt</i> files specify the timing of each detected leaning event, enabling precise alignment of behavioral labels with sensor data. All data has been thoroughly anonymized, cleaned, and checked for completeness.This dataset enables research on automated pose estimation, behavioral analysis, and social signal processing in VR, with potential applications in soft-skills training, accessibility, and human-computer interaction; and facilitates new investigations into nonverbal behaviors such as leaning, mirroring, and engagement during simulated interviews.For detailed information on data structure, file formats, and usage, please see the included <i>README.md</i>.