Biomechanics-informed inertial tracking achieves the accuracy of marker-based kinematics

Abstract

Inertial measurement units (IMUs) could transform human movement science by enabling motion tracking outside the laboratory. Yet, their accuracy is perceived as inferior to that of marker-based systems. Here, we introduce IMoveLab, a method that integrates biomechanical priors into state-estimation filters, harnessing coupling across degrees of freedom and the spring-like behavior of muscle–tendon units for recursive state correction. We found that modern state-estimation filters match the accuracy of marker-based knee tracking when evaluated against biplane radiography, the gold standard for skeletal motion. IMoveLab further improves accuracy and achieves drift-free lower-extremity kinematics in long-duration trials, whereas alternative approaches do not. Movement scientists have been anticipating the leap out of the laboratory with cautious optimism. With methods and evidence that support more confident adoption of IMUs, this leap seems less like an aspiration and more like the first line of a new chapter written in the environments where human movement unfolds. Wearables could bring movement analysis into everyday life but concerns about accuracy over long recordings limited their adoption. Here, the authors show that inertial tracking can match the accuracy of marker-based motion capture and present a biomechanics-informed sensor-fusion approach that eliminates drift over long durations.