Predicting Athlete Ground Reaction Forces and Moments From Spatio-temporal Driven CNN Models
Conventional methods to generate biomechanical data, required for traditional inverse dynamics estimation of athlete joint forces and loads, are confined to biomechanics laboratories far removed from the sporting field of play. This study used deep learning to predict 3D ground reaction forces and moments (GRF/M) from legacy marker-based motion capture sidestepping trials, ranking correspondence of multivariate regression from five convolutional neural network (CNN) models against ground truth force plate data. By fine-tuning from CaffeNet, a model derivative of ImageNet, mean predicted GRF/M correlations to ground truth above 0.97 were achieved for complex sport-related movements.
Neural Data-Driven Musculoskeletal Modeling for Personalized Neurorehabilitation Technologies
This review aims to discuss clinically viable methods for accessing the neural information underlying an individual’s movement from electrophysiological recordings and the development of subject-specific musculoskeletal modeling formulations that can be driven by the extracted neural features.