Design of a subject-specific EMG model for rehabilitation movement

In the latest years, robotics technologies have been increasingly introduced in rehabilitation with the objective of cost reduction and speed up the recovery process. While with most of the devices the patient is passive, the current challenge is to build devices able to understand patient’s intention and adapt accordingly, forcing his/her active involvement. A way to understand the patient is to use EMG-driven neuromusculoskeletal model able to compute muscle dynamics and joint torques from the electromyography (EMG) signals. While the approach is quite promising, collecting EMG data is still not a simple task as placement of electrodes requires professional skills and EMG data can be affected by electric and magnetic noise. This work proposes a model that builds upon a reduced experimental database of EMG data from a common rehabilitation movement to develop the capability of predicting EMG values for the same movement executed at arbitrary speed. The reported experimental results are promising, showing a good accuracy in EMG prediction thus enabling the possibility of their use as input for EMG-driven neuromusculoskeletal models. Model applicability, even if limited to repetitive movements, can simplify the use of active rehabilitation devices and still keeping their possibility to be driven by patient.