Making wearable motion estimates trustworthy: validation against optical ground truth, calibrated per-joint uncertainty, and robustness in free-living use. Turning each kinematic estimate into a signal that is verificated, carries its own confidence, and detects its own failures.
Objectives:
Metrological validation of wearable estimates against optical motion capture as ground truth.
Calibrated, per-joint uncertainty quantification (conformal and Bayesian methods).
Sensor-displacement and misplacement detection, and drift handling, for long-term free-living use.
Out-of-distribution detection and reliability flags that make estimates auditable for clinical and sports decisions.
Reproducible validation protocols; validation datasets released openly (data papers, DOI) as a by-product of specific studies.
