Multidimensional Ground Reaction Forces and Moments From Wearable Sensor Accelerations via Deep Learning

Objective: Monitoring athlete internal workload exposure, including prevention of catastrophic non-contact knee injuries, relies on the existence a custom early-warning detection system. This system must be able to estimate accurate, reliable, and valid musculoskeletal joint loads, for sporting maneuvers in near real-time during match play. However, current methods are constrained laboratory instrumentation, labor cost intensive, require highly trained specialist knowledge, thereby limiting their ecological validity wider deployment. An informative next step towards this goal would new method obtain ground kinetics field. Methods: Here we show that kinematic data obtained from wearable sensor accelerometers, lieu embedded force platforms, can leverage recent supervised learning techniques predict multidimensional reaction forces moments (GRF/M). Competing convolutional neural network (CNN) deep models were using laboratory-derived stance phase GRF/M simulated accelerations running sidestepping derived nearly half million legacy motion trials. Then, predictions made each model driven by five recorded independent inter-laboratory capture sessions. Results: The proposed workbench achieved correlations truth, maximum discrete GRF component, vertical F z 0.97, anterior Fy 0.96 (both running), lateral Fx 0.87 (sidestepping), with strongest mean across components 0.89, GRM 0.65 sidestepping). Conclusion: These best-case indicate plausibility approach although range results was disappointing. accurately on-field will improved lessons learned study. Significance: Coaching, medical, allied health staff could ultimately use technology monitor loading indicators game play, aim minimize occurrence injuries elite community-level sports.