Accuracy of 3D Motion Data During Dynamic Movements for Head, Controller, and Hand Tracking Estimated with a Head-Mounted Display

This study assessed the spatial tracking accuracy of a head mounted display (HMD) under simulated dynamic exercise and gaming conditions, using both controller-based and markerless hand tracking inputs. The aim was to evaluate the HMD’s suitability for clinical and research applications by investigating three key aspects: (1) overall tracking accuracy, (2) the influence of different movement conditions, and (3) the effect of recording duration on spatial accuracy. Ten healthy participants completed six distinct virtual scenarios designed to elicit a broad range of head, arm, and full-body movements, including both highly dynamic locomotor tasks and slower, localized body movements. More than 225 minutes of data were recorded simultaneously with a state-of-the-art marker-based motion capture system and the HMD. Root mean square errors (RMSE) between the trajectories of both systems served as the main outcomes. Head tracking showed high accuracy across all tasks, with overall RMSE values below 4 mm. Controller-based hand tracking yielded RMSEs of 11.8 mm (left) and 8.3 mm (right). In contrast, markerless hand tracking resulted in higher errors (approx. 37-40 mm), limiting its use in precision-critical settings. Movement complexity impacted tracking performance. Head and controller errors remained low in steady tasks but increased with dynamic or occlusion-prone motions. Additionally, time-dependent drift was observed, with HMD tracking accuracy degrading slightly over nine-minute intervals. These findings underscore the potential of HMD-based tracking in digital health, while highlighting the need for further improvements in drift correction and markerless hand tracking precision for broader clinical adoption.