Musculoskeletal Modeling of Human Locomotion Based on Low-Dimensional Impulsive Activation Signals: Perspectives for Neurotechnologies

Abstract

Human locomotion can be expressed as the action of impulsive activation signals over muscle groups. This view has been used in this study for generating a musculoskeletal model for the prediction of muscle forces and joint moments. A set of activation signals of impulsive nature has been extracted from experimental electromyographic (EMG) recordings in several locomotor tasks. These activations signals were used as input to the musculoskeletal model whose unknown parameters were obtained by calibration. Once calibrated, the model could work in open-loop, estimating joint moments over multiple degrees of freedom using only the recorded kinematics (and the internal impulsive controller). It is shown that the accuracy in estimation of the joint moments was comparable when using the low-dimensional activations signals, as proposed in this study, with respect to using the experimental EMG signals, as in more common EMG-driven musculoskeletal models. These results have implications, which are discussed in this contribution, in the design of control systems in human-machine interfacing for neurorehabilitation, such as active orthoses or prostheses.

This work was supported by the ERC Advanced Grant DEMOVE (to DF).