Estimation Of Joint Angle From Ground Reaction Force In Human Gait

Current research in human biomechanics is trending towards machine learning based gait parameter estimations. This work is a proof of concept in using Artificial Neural Networks (ANNs) to estimate lower limb kinematics from foot kinetics. In this study, we present a three-layer Feed Forward Neural Network (FFNN) to estimate ankle angles from Ground Reaction Forces (GRFs). GRFs are measured from instrumented foot insoles (MOTICON) while ankle angles are measured using an optical motion camera system. Salient input features and target outputs for the ANN are selected based on priori knowledge of five gait event occurrences at its pre-defined gait intervals. These five main gait events are; Heel Strike (HS), Foot Flat (FF), Mid Stance (MS), Heel Off (HO) and Toe Off (TO). Results indicate high correlations between estimated and its ground truth angles (NRMSE <; 9.112% and ρ > 0.94). The result from this study shows the possibilities of modelling a dual-purpose foot insole that can measure GRFs while estimating lower limb angles. This will open up countless opportunities for outdoor gait monitoring during acts of daily living.