Socially Assistive Robotics (SAR) has shown to be an important tool to assist patients in physical rehabilitation. SAR is used to provide feedback about patient's state and performance to users and health professionals, therefore, patients are monitored by means of sensor interfaces. In this context, aiming to avoid over-training conditions, one of the most important parameter to monitor is the fatigue level. However, it is usually measured by subjective scales such as Borg scale, thus, there is a need to develop systems that are able to estimate fatigue with greater accuracy. It has been demonstrated that fatigue can be associated to the decreasing performance of the exercise. Hence, this work carried out a study to determinate which temporal and kinematic features are related to the fatigue level during a sit-to-stand test. The procedure consisted of sitting down and standing up from a chair while kinematic data were measured by a Kinect sensor, in order to relate kinematic data and fatigue. Results show that temporal features (time stand-to-stand and time sit-to-stand) and 3 kinematic features (max vertical-velocity of the spin base, max knee-flexo-extension velocity, and max hip-flexo-extension velocity), have a significant correlation with the fatigue level $(p \lt 0.05)$.