Phase-Separated Complex Hilbert PCA on Markerless 3D Pose Estimation Data: A Global Phase Network and Its Extension to a Continuous Field on the Body Surface

Researchers from Japan have developed a novel framework that applies Complex Hilbert Principal Component Analysis (CHPCA) to markerless 3D pose estimation data, extracting a single dominant whole-body phase pattern as a complex eigenvector. The method, detailed in a preprint on arXiv, separates motion into distinct phases (backswing and downswing) automatically without priors on strike count or rest location. It further extends the analysis to 1,079 body-surface mesh vertices, representing the kinematic chain as a continuous phase field across the body.

On 14 hammer-striking trials from a single subject, the framework revealed a trunk-anchored global phase architecture and a functional asymmetry between preparation and execution phases. Mode-1 contribution was 45.5% in the backswing versus 70.5% in the downswing, with inter-trial Spearman consistency of 0.38 vs 0.58. A correspondence analysis showed strong positive correlation between Mode-1 amplitude and kinetic-energy mobilisation variance in the downswing (ρ ≈ 0.71 on both skeleton and mesh) and no correlation in the backswing, indicating the framework bridges kinematic and kinetic coordination descriptions.