Prediction of ground reaction forces and moments and joint kinematics and kinetics by inertial measurement units using 3D forward dynamics model

• Published in: Journal of Biomechanical Science and Engineering 2024, Vol.19(1), pp.23-00130-23-00130
• Main Authors: HARAGUCHI, Naoto, HASE, Kazunori
• Format: Article
• Language: English
• Subjects: Accuracy; Ankle; Biomechanical analysis; Biomechanical engineering; Biomechanics; Contact model; Cost function; Current prediction; Force plates; Gait; Gait analysis; Hip; Human model; Human motion; Inertial platforms; Joints (anatomy); Kinematics; Kinetics; Mathematical models; Mechanical properties; Motion capture; Movement; Multibody dynamics; Optimization; Predictions; Statistical models; Therapeutic applications; Three dimensional motion; Torque
• ISSN: 1880-9863; 1880-9863
• DOI: 10.1299/jbse.23-00130

Predicting the ground reaction force (GRF) and ground reaction moment (GRM) with a biomechanical model-based approach has an advantage for biomechanical gait analysis in situations where a statistical model cannot be used due to a lack of training data. However, the current prediction methods using a biomechanical model have some issues for clinical application. The present study developed a new inertial measurement unit (IMU)-based method to predict the GRF, the GRM, and the joint kinematics and kinetics with a 3D biomechanical model and simple system. The present method predicts them using a 3D forward dynamics model that computationally generates human movements that minimize the hybrid cost function defined by physical loads and errors between the motion of the model and that of the participants recorded by six IMUs, which allows the prediction system to use only a relatively small number of IMUs. We investigated the prediction accuracy during walking by comparing the new method with a conventional analysis using a force plate and motion capture system. As a result, we observed strong and excellent correlations between the prediction and measurement of the anterior GRF, vertical GRF, sagittal GRM, hip flexion angle, knee flexion angle, hip flexion torque, and ankle dorsiflexion torque. Considering the accuracy of previous studies and that required for gait analysis, the present method could predict them with practical accuracy due to estimated biomechanically valid motions based on optimization using the hybrid cost function that includes a biomechanical evaluation. Moreover, the prediction system has an advantage for clinical applications because the present method observed practical accuracy that has the potential to be applied to some sports analysis and can analyze 3D motion with a simple system consisting of a small number of hardware components and a software.