Markerless motion capture systems, such as Theia3D, offer a promising alternative to traditional marker-based systems (e.g., Vicon) not only for dynamic motion analysis, but also potentially for postural control research. The purpose of this study is to simultaneously derive and then compare principal movements (PMs) from Theia3D and Vicon data collected from 13 volunteers during eyes-open and eyes-closed bipedal quiet stance using principal component analysis (PCA). Agreement between systems was assessed by computing correlation coefficients for corresponding PM time series and evaluating their sensitivity to visual condition changes. Theia3D and Vicon produced highly similar PM structures, with the first two PMs—representing anterior-posterior ankle-strategy and medio-lateral sway—exhibiting near-perfect correlation (r = 99.8 % and r = 99.0 %) and explaining a comparable portion of the entire postural variance (PM1: 74 % vs. 76 %; PM2: 9.4 % vs. 9.9 % for Vicon and Theia3D results, respectively). The first seven PMs, which collectively accounted for 97 % of postural variance, showed strong agreement (r = 0.814 to 0.928) despite fundamental differences in marker placement and tracking methodology. Correlation declined in higher-order PMs, suggesting they capture distinct aspects of postural movements or system-specific noise. Both systems demonstrated comparable sensitivity to postural control modulations induced by visual deprivation. Notably, Theia3D exhibited higher noise levels, particularly in higher-order PMs, indicating that pre-processing with filtering should be considered before PCA. Overall, these findings confirm that Theia3D provides highly comparable results to Vicon for postural PMs, reinforcing its potential as a valid and efficient alternative for postural control research using PCA-based approaches.
Keywords:
AI-based motion tracking; Markerless motion capture; Principal component analysis (PCA); Principal movements (PMs); Postural control