Towards accurate 3D analysis of human movement: An integrated UWB/MIMU approach

Accurate 3D analysis of human movement (3D AHM) is essential for diagnosing movement disorders, guiding therapeutic interventions, and monitoring rehabilitation outcomes. Although optical motion capture remains the gold standard for biomechanical assessment, its reliance on expensive laboratory infrastructure and time-intensive workflows restricts its use in routine clinical practice. This limitation has driven the demand for portable, objective, and clinically viable alternatives capable of providing reliable movement data outside controlled lab environments. Wearable sensing technologies, particularly Magnetic Inertial Measurement Units (MIMUs), offer promising opportunities for ambulatory movement analysis. While MIMUs accurately capture segmental orientations, deriving precise positional information remains difficult due to drift from double integration of acceleration signals. Many clinically meaningful gait and balance metrics depend on accurate segmental positions, underscoring the need for improved positional accuracy. One solution is to fuse complementary sensing modalities. Ultra-Wideband (UWB), with its drift-free distance measurements, presents a strong candidate for this purpose.

This PhD dissertation investigates the feasibility of an integrated UWB/MIMU (UMIMU) wearable system for ambulatory 3D AHM, with a focus on improving positional accuracy and long-term stability. The research progressed through five stages. First, a comprehensive literature review identified major gaps in existing research, including insufficient characterization of UWB performance on the human body, lack of integrated platforms, and inadequate positional accuracy for clinical use. Second, a custom hardware system combining UWB and MIMU sensing was developed and experimentally characterized under both line-of-sight (LOS) and human-induced non-line-of-sight (NLOS) conditions. Third, a novel swarm-based calibration method was introduced, reducing systematic UWB ranging errors to sub-centimeter levels. Fourth, a swarm-optimization-based position estimation algorithm was developed and validated using synthetic UWB data. Finally, a UMIMU fusion algorithm based on an Extended Kalman Filter framework was created and validated with synthetically generated data. The integrated UMIMU system achieved mean position estimation errors of approximately 6 cm (SD ≈ 3 cm), demonstrating strong consistency and robustness for continuous monitoring applications. While not yet a replacement for optical motion capture, the proposed methods establish a solid technological foundation for future clinically deployable wearable systems capable of delivering reliable 3D AHM outside laboratory settings.