Estimating EMG signals to drive neuromusculoskeletal models in cyclic rehabilitation movements.

A main challenge in the development of robotic rehabilitation devices is how to understand patient's intentions and adapt to his/her current neuro-physiological capabilities. A promising approach is the use of electromyographic (EMG) signals which reflect the actual activation of the muscles during the movement and, thus, are a direct representation of user's movement intention. However, EMGs acquisition is a complex procedure, requiring trained therapists and, therefore, solutions based on EMG signals are not easily integrable in devices for home-rehabilitation. This work investigates the effectiveness of a subject- and task-specific EMG model in estimating EMG signals in cyclic plantar-dorsiflexion movements. Then, the outputs of this model are used to drive CEINMS toolbox, a state-of-the-art EMG-driven neuromusculoskeletal model able to predict joint torques and muscle forces. Preliminary results show that the proposed methodology preserves the accuracy of the estimates values.