Using EMG Data of Reachable Muscles to Estimate the Activation of other Muscles During Shoulder Press Movement

Abstract

Knowing the required muscle activation pattern of a determined movement can be used as an input to functional electrical stimulation in order to artificially activate the involving muscles in individuals with paralyzed limbs. Although there are muscles that are far from the skin, EMG data acquisition cannot be done noninvasively. There are several studies that estimate the muscle activations using the kinematics of the motion. The measurement devices for the joint angles can be volume occupying and may limit the dexterity of the motion. This article proposes to predict the missing anterior deltoid activation using the sEMG data of the long head and lateral head of triceps during shoulder press movement. First, the joint angles of the shoulder and elbow are estimated applying extended Kalman filter on an EMG-based state-space model. Having the kinematics of the motion, the joint torques can be determined using upper arm musculoskeletal model and inverse dynamics controller to track the estimated the joint angles. A static optimization method and Hill-based model are applied so the muscle activation of the muscles can be determined. An experimental setup is designed to obtain the biological and kinematic data needed to construct the equations, and the real values of the angle and activations can be used for the validation of this method. The RMSE of the real and estimated anterior deltoid activation is between 0.15 and 0.21 that is acceptable.