PhD Defence Vinish Yogesh | Towards accurate 3D Analysis of Human Movement | An integrated UWB/MIMU approach

Towards accurate 3D Analysis of Human Movement | An integrated UWB/MIMU approach

Vinish Yogesh is a PhD student in the department Biomedical Signals and Systems. (Co)Promotors are prof.dr. J.H. Buurke and dr.ir. B.J.F. Beijnum from the faculty of Electrical Engineering, Mathematics and Computer Science (EEMCS), University of Twente.

Accurate 3D Analysis of Human Movement (3D AHM) is fundamental in clinical rehabilitation for understanding movement disorders, guiding interventions, and monitoring recovery. Traditional optical motion capture systems remain the gold standard for biomechanical analysis due to their sub-centimeter accuracy, yet their reliance on fixed laboratory setups, high cost, and time-intensive data processing limits their use in routine clinical practice. Current clinical assessments, therefore, rely heavily on visual observations and broad functional tests, highlighting the need for objective, portable, and cost-effective movement analysis systems that can operate outside the lab.

Wearable sensor technologies, particularly those based on Magnetic Inertial Measurement Units (MIMUs), offer a promising alternative. MIMUs have matured in their ability to capture angular kinematics, but accurate position estimation remains a challenge due to drift from the double integration of acceleration signals. Many clinically meaningful parameters, such as step length, base of support, and center of mass trajectory, require accurate segmental positions rather than only angular orientations. While algorithmic improvements have reduced drift, progress is nearing its practical limit. Integrating complementary sensing modalities that provide drift-free positional information, such as Ultra-Wideband (UWB), represents a promising direction.

This PhD research therefore investigated the feasibility and potential of an integrated UWB/MIMU wearable sensor system (UMIMU) for ambulatory 3D AHM, specifically focusing on improving position estimation accuracy and stability. The work progressed through five stages: (1) a comprehensive literature review identifying the gap between existing UWB/MIMU research and the needs of clinical movement analysis. This review identified methodological gaps such as limited characterization of UWB performance on the human body, lack of integrated hardware platforms, and insufficient accuracy for clinical use, thereby defining the research needs and design requirements for the subsequent studies. (2) A fully integrated UWB/MIMU hardware platform was developed, and systematic experimental characterization of UWB behaviour under line-of-sight (LOS) and human body-induced non-line-of-sight (NLOS) conditions was studied. The findings confirmed the need for additional calibration and specialized algorithms tailored for short-range, on-body use. (3) To address the limitations, a novel swarm-based calibration method was introduced to address the systematic errors in UWB distance estimates. This calibration reduced the systematic ranging errors to sub-centimeter levels. (4) Building further, a swarm optimization-based position estimation algorithm was developed for estimating positions from UWB ranging data and was validated with synthetic UWB data derived from motion capture trajectories. (5) Finally, a UMIMU fusion algorithm was developed based on an Extended Kalman Filter (EKF) framework combining MIMU and UWB data to achieve accurate and drift-free position estimates. Using synthetically generated UWB distance estimates and measured MIMU data, the proposed UMIMU fusion algorithm was validated across varying levels of bias and random errors in distance estimation.

The integrated sensor system achieved mean position estimation errors around 6 cm (SD ≈ 3 cm), demonstrating strong consistency and drift resistance, an essential feature for continuous monitoring applications in rehabilitation, fall-risk assessment, and home-based care. Though the achieved accuracy does not yet reach the sub-centimeter benchmark of optical systems, the system’s portability, stability, and adaptability make it a viable foundation for future clinical translation. The developed calibration, position estimation, and data fusion methods are also generalizable to other short-range ranging technologies.

Overall, this PhD dissertation supports the feasibility of an integrated, portable UWB/MIMU sensor system for clinical 3D AHM, with consistent, drift-free position estimates. While it does not yet offer a ready-made replacement for laboratory-based 3D AHM, this study represents a significant step toward the actual implementation of wearable sensor technology in clinical practice by establishing a technological foundation for future clinically deployable systems.