The 3D CoM kinematic estimation using a simple machine learning for portable gait monitoring

Abstract

The center of mass (CoM) can be used as a major factor for gait monitoring. Therefore, many studies have tried to estimate the CoM kinematics using an inertial measurement unit (IMU) attached to the sacrum, but the estimation accuracy of velocity and acceleration was not high enough. In this study, we estimated the 3-axis CoM kinematics by first estimating CoM velocity using a machine learning. The change in velocity was estimated from the sacral IMU data using an artificial neural network (ANN), and the position and acceleration were derived by integrating and differentiating the velocity. We estimated the CoM kinematics with high accuracy, which reduced an error of up to 4 times compared to the sacral kinematics. With help of this accurate estimation method, it is expected that the CoM-based gait monitoring with a single IMU would be improved.