Deep Learning-Based Ground Reaction Force Estimation for Real-Time Clinical Applications

This study represents a significant step forward in real-time, marker-based gait analysis, offering a cost-effective and efficient method for estimating Ground Reaction Forces (GRF) and Center of Pressure (COP) in clinical settings. Using marker trajectories captured by a motion capture system, we trained a recurrent neural network (RNN) to estimate GRF and COP during treadmill walking. Real-time experiments with five healthy subjects demonstrated high to very high correlations between predictions and ground truth (r_COP =0.81, r_GRF = 0.96). These promising results highlight the potential of this approach to transform clinical gait analysis by reducing reliance on expensive and fixed force-plate systems. Future work will focus on validating the method with pathological gait datasets to ensure broader applicability and advancing its integration into diverse clinical environments for real-time feedback and analysis.