A Neuromechanical Model of Knee Flexion-Extension Based on Linear Regressors for Neuro-Controlled Exoskeletons

Abstract

In this paper we evaluate the use of linear regressors to reconstruct knee angles from electromyographic signals recorded from lower-limb muscles in charge of knee flexion and extension. This neuromechanical model has been tested under two different speed conditions showing a very reliable reconstruction of knee kinematics for all participants. A generalized model, including all participants, has also been tested with a small reduction of decoding accuracy, but still showing enough reliability. The application of neuromechanical modelling has important advantages in the generation of a more natural and embodied control of lower-limb exoskeletons. The proposed model is expected to have a robust behavior in a real-time control of this kind of devices.