Estimation of Muscle Torques from EMG and Kinematics During Planar Arm Movements

Estimating the contributions of individual muscles to limb movements is crucial to our understanding of the organization of the motor system. In pathological conditions, characterizing the roles of the individual muscles may provide a basis for devising personalized treatments. Recently, myoproces-sors have been used to computationally alter muscles directions of action (`virtual biomechanics') in order to study the motor system's modular structure and adaptive capabilities in isometric conditions. As a step toward extending these studies to actual movements, here we describe a myoprocessor model which is capable for predicting the muscle forces during planar arm movements in real-time, from EMG and movement data. We use a polynomial model to account for multi-joint muscle geometry and a Hill-type muscle model to predict muscle forces during movements. We report on an experimental apparatus and a two-step parameter estimation procedure which involves both isometric force generation and actual movements. Preliminar results show a good reconstruction performance in isometric conditions. During movements, the reconstruction performance generally degradates, specially for shoulder torque, and is only acceptable for some movements.