Neuromuscular Control with Forward Dynamic Approximation in Human Arm
Jan Thomas, Vinupritha P and Kathirvelu DDepartment of Biomedical Engineering, SRM University, Kattankulathur–603 203, Tamilnadu, India.
Corresponding Author E-mail: vinupritha@gmail.com
Abstract: The central nervous system (CNS) directs a large number of muscles to produce complex motor behavior’s. Moreover, human movement control is significantly compromised in neuromuscular diseases; many of which result from the imbalance in the sensorimotor control system. Forward models are also known to capture the casual relationship between the inputs to the system and its output. The aim of this study was to approximate a forward dynamic simulation by using feed forward neural networks in human elbow arm movement. Motion capture data was used to generate c3d data from Qualisys software via Qualisys track manager. The c3d data was then converted into marker data which contains the markers location in the form of x-y-z coordinate. OpenSim- biomechanical software was used to process the marker data for scaling, inverse kinematics and the subsequent forward dynamics simulation in the right human upper arm model. The results for the training error of the approximated forward dynamics simulation for 3 degrees of freedom (DOF) were 2.804429, 1.468017 and 2.475500 with a constant validation error of 0.000000. The proposed control algorithm served as robust method for approximating a forward dynamic simulation, and can broaden the understanding the representations of neuromuscular control in the central nervous system as representations of movement plans that are eventually executed by the spinal cord and muscles in the periphery.
Keywords: Neural networks; Open Sim; Forward dynamics; CNS Back to TOC