Physical calibration and signal periodic updates to synchronize IMUs for 5-day continuous knee flexion tracking
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.
Authors
- Ke Song (ORCID: https://orcid.org/0000-0001-8987-2523)
- Josh R. Baxter (ORCID: https://orcid.org/0000-0003-0269-2170)
Institutions
- University of Pennsylvania (US)
Publication Details
- Journal
- medRxiv
- Published
- 2026-11-16
- DOI
- https://doi.org/10.64898/2026.11.14.26368098
- Primary Topic
- Inertial Sensor and Navigation
- Type
- preprint