In this paper, we present MyoTrack, a two-step classification routine that uses a wearable surface electromyography (sEMG) sensor to identify the level of subject participation during robot assisted rehabilitation. In the first step, we use sEMG activation as a measure of patient participation; stating that, high sEMG correlates with high participation. We then extract the subject's hand trajectory using the Myo's inertial measurement unit. The hand trajectory is compared with the robot's trajectory to identify whether the high muscle activity observed is due to the active participation of the subject in therapy or due to random gestures and motions. As the robot assistance considered in this paper can autonomously complete the therapy task without any subject participation, it is crucial to identify this participation level. We demonstrate that high muscle activation along with high similarity between the hand and robot end-effector trajectory is a reliable indicator of subject participation with an accuracy > 90% for the cases considered.