Volitional PD computed torque control design of a 2-DOF finger model for cylindrical grip movement assistance with sEMG signal classification

In this paper, we present a two-link finger model design for a hand exoskeleton that performs cylindrical grip movements. Movement intention was initially detected with sEMG signal classification from a subset of the Ninaweb dataset. Then, a simulated PD Computed Torque Control (PD CTC) allowed for accurate movements given a defined intention. A high accuracy of classification was obtained using a single random forest classifier and a very small error was obtained for the fingers when following the desired trajectories.