pipette
ENEnglish

Physical calibration and signal periodic updates to synchronize IMUs for 5-day continuous knee flexion tracking

K. Song, J. R. Baxter

PreprintAfirmaciones fuertes, leer con cuidadoUso en el mundo real

En palabras de los autores

Wearable inertial measurement unit (IMU) sensors provide a promising method to assess real-world human motion kinematics. However, multi-day multi-IMU tracking is difficult due to the desynchronization driven by inaccurate IMU clocks. In this study, 15 volunteers wore IMUs continuously on their thighs and shanks for 5 weekdays (Mon-Fri), which included 3 repeated in-lab motion capture sessions. We synchronized thigh and shank IMUs using two methods. The first was physical calibration, where we maneuvered the IMUs in a rigid holder to produce sync signals pre-donning (Mon) and post-doffing (Fri), then linearly resampled data from one IMU to match the other via cross-correlation. The second was signal-based periodic updates, where we shifted IMU clocks hourly based on the timewise alignment of preprocessed body-worn acceleration signals. After syncing, we used sensor-to-segment alignment and accelerometer-gyroscope fusion to estimate knee flexion and compared results to camera-based motion capture. Physical calibration alone removed clock offset in most cases, where IMU-estimated knee flexion generally matched motion capture (correlation [≥]0.944; root-mean-square difference RMSD [≤]6.9 degrees). But in some participants, residual mid-week desynchronization (~0.2s) resulted in misestimated knee flexion (Wed: RMSD up to ~19 degrees). After applying signal-based updates or combining it with physical calibration, IMU matched motion capture on all 3 days (Mon: r [≥]0.975, RMSD [≤]6.1 degrees; Wed: r [≥]0.971, RMSD [≤]6.2 degrees, Fri: r [≥]0.977; RMSD [≤]5.9 degrees). Our multi-modal synchronization enables week-long knee flexion tracking and lays a foundation for advanced syncing tools, for improving the feasibility of wearable multi-IMU systems to continuously monitor real-world knee kinematics in clinical studies.

Resultado principalLimitación que admiten los autores

Apareció: martes, 17 de noviembre. medRxiv. Preprint, todavía sin revisión por pares.

DOI: 10.64898/2026.11.14.26368098