IMU-Based Posture Monitoring for Rehabilitation Applications

Abstract

Accurate assessment of posture is important in rehabilitation-related applications, where abnormal body alignment may contribute to musculoskeletal discomfort, altered movement patterns, and reduced functional performance. Inertial Measurement Units (IMUs) provide a portable and low-cost alternative to laboratory-based motion capture systems for posture monitoring. This thesis investigated the accuracy and reliability of IMU-based posture estimation using Movella DOT sensors, with a Qualisys optical motion capture (MoCap) system used as the reference measurement. Four sagittal posture variables were evaluated: pelvic tilt, thorax tilt, thorax–pelvis relative alignment, and forward head posture (FHP). The influence of breathing on thorax-mounted IMU measurements was investigated, and the feasibility of IMU-based respiration monitoring was also explored. The results demonstrated good agreement between IMU-based posture estimates and the MoCap reference for pelvic tilt, thorax tilt, and thorax–pelvis alignment, with mean RMSE values of 0.86◦, 1.65◦, and 1.82◦, respectively, and ICC values above 0.96 for all three variables. Forward head posture showed lower agreement and greater inter-subject variability. The comparison between Euler-angle–based and quaternion-based methods revealed only minor differences under the evaluated posture conditions. Deep breathing introduced visible oscillations and increased variability in thorax-mounted IMU signals, although overall IMU–MoCap agreement remained high. IMU-based respiration monitoring demonstrated moderate to strong agreement with a respiratory belt reference during deep breathing, but lower reliability during normal breathing. Overall, the findings indicate that IMU-based sagittal posture estimation is feasible for rehabilitation-related posture monitoring, particularly for pelvic and thoracic measurements.