Dynamic characteristics of human gait helps to resolve the trade-off between the monitoring performance and the simplicity of wearables

Wearable devices enables us to monitor human walking in the field. There is however technical trade off exists between the data richness and the wearable convenience. To extract motion information from the limited measurement data, we propose to exploit the biomechanical characteristics of human gait describable by the spring mechanics. The unmeasured ground reaction forces (GRF), joint torques, joint kinematics were estimated from a single measurement of the center of mass (CoM) using a compliant legged walking model, which is also used to improve the accuracy of GRF estimation from machine learning despite the small number of training data set.Results imply that the use of biomechanical characteristics of human gait helped to resolve the trade-off between the monitoring performance and the simplicity of wearables.