Author(s): TAN, K., KING, M., MCERLAIN-NAYLOR, S., Institution: LOUGHBOROUGH UNIVERSITY, Country: UNITED KINGDOM, Abstract-ID: 2700

INTRODUCTION:
Objective quantification of joint kinematics is central to understanding movement technique and injury mechanisms in sport. While optical motion capture remains the reference standard for joint angle estimation, its application is largely confined to laboratory environments. Inertial measurement units (IMUs) offer a practical alternative for field-based assessment, but their validity has been predominantly evaluated in linear and repetitive tasks. Many sport-specific movements, however, are non-linear and highly variable. Badminton lunges represent a demanding example, involving rapid changes in segment orientation and large joint excursions across the knee and ankle. Establishing the accuracy of IMU-derived joint angles under such conditions is a necessary step toward extending biomechanical analysis beyond the laboratory.
METHODS:
Thirty-six university-level badminton players (age: 22 ± 2 years) performed badminton-specific forward lunges while kinematic data were collected using optical motion capture and body-worn IMUs. Marker-based joint angles were obtained using a standard OpenSim inverse kinematics workflow, and IMU-derived joint angles were computed using an OpenSense-based pipeline within OpenSim. OpenCap camera-based approach was used to generate a musculoskeletal model, which served as the common model input for both analyses. Knee and ankle joint angles were extracted with emphasis on sagittal-plane motion. Kinematic waveforms were time-normalised to stance and compared using root mean square difference (RMSD) and an agreement statistic (intraclass correlation coefficient, ICC). Ethical approval was obtained from the Loughborough University ethics committee, and all participants provided informed consent
RESULTS:
IMU-derived knee flexion–extension angles closely followed marker-based waveforms across stance, with consistent timing of key movement features. Knee angle RMSD was 8.2 ± 4.0°, with larger discrepancies during early and mid-stance and reduced differences toward push-off. Agreement between systems remained high across most of stance, with an intraclass correlation coefficient indicating strong waveform similarity across repeated trials (ICC (2,1).
At the ankle, dorsiflexion–plantarflexion waveforms showed greater variability, with RMSD of 1.4 ± 0.3° and the largest discrepancies at initial contact and early stance.
CONCLUSION:
Agreement between IMU and marker-derived joint angles was joint and phase dependent during this non-linear badminton lunge task, with more robust performance observed at the knee than the ankle. IMUs preserved key features of lower-limb kinematic waveforms, but systematic differences in magnitude were evident during rapid movement phases. The preserved waveform similarity supports the use of wearable IMUs for applied movement analysis in sport, provided joint and phase-specific limitations are recognised (RMSD ~8° at the knee and ~1° at the ankle), and represents an important step toward wearable biomechanical assessment pipelines.