In this project, we present MyoTrack, a realtime classification procedure that identifies levels of subject participation during robot-assisted rehabilitation using a wearable surface electromyography (sEMG) Myo armband. We hypothesize, test, and prove that high sEMG correlates with high participation and vice versa. We then use Myo's inertial measurement unit to extract the subject's hand trajectory during the rehabilitation task. Comparing this hand trajectory with the ground-truth robot trajectory enables us to identify whether any high muscle activity corresponds to the active participation of the subject in robotic therapy or not (random gestures and motions). Since the robotic assistance implemented in this paper can autonomously complete the therapy task without any subject participation; it is crucial to identify the patient's participation level in realtime and develop a suitable intervention strategy. Using 15 healthy subjects, we demonstrate that the proposed methodology of combining sEMG activation and robot-hand trajectory matching is a reliable indicator of subject participation with a realtime accuracy of 91.45%. We also present a realtime application that uses MyoTrack in the back-end to identify the realtime participation of a subject and intervene accordingly.