Hybrid hill-type and reflex neuronal system muscle model improves isometric EMG-driven force estimation for low contraction levels

Abstract

Non-invasive muscle force estimation by EMG signals can be obtained using a priori information provided by mathematical models of muscle dynamics. In this study, EMG-driven isometric force estimates performed using three muscle model formulations are compared to verify contribution of stretch reflex to muscle force estimation: (1) Winters’ Hill-type model with muscle spindles and Golgi tendon organ reflex system; (2) Winters’ model without reflex; and (3) Zajac model adapted by Menegaldo (ZM model). Submaximal isometric plantar flexion torque predictions, estimated by the models mentioned above, were compared with dynamometer measurements. Surface EMG was collected from gastrocnemius medialis, gastrocnemius lateralis, soleus, and tibialis anterior muscles from 12 volunteers and synchronized with dynamometer plantar flexion torque measurements. The experimental protocol consisted of sustained contraction intensities of 20 and 60% of individual maximum voluntary contraction (MVC) assisted by real-time visual feedback. The results show the improvement of torque prediction accuracy for the reflex model (1) at 20% MVC, which was not observed for 60% MVC.