Increasing movement frequency and patient engagement is a crucial part of the rehabilitation process. One approach for integrating productive rehabilitation tasks with motivational features is through an interactive game. Furthermore, if moving the impaired body part can generate electricity to power the system it could motivate the patient even more. Therefore, in this work we developed a wearable elbow rehabilitation system using self-powered sensors based on the triboelectric nanogenerator. The system consisted of a 3D printed body integrated with the self-powered sensors that could provide real-time elbow joint motion data. After digitizing the acquired data, the elbow motion could manipulate objects in an interactive game made in Python language. We first provide detailed analysis of the system design and the triboelectric nanogenerator output which was enhanced using gears and material surface nano-patterning. Next we describe the sensor data processing circuit and game control using Arduino, based on the direction and extent of forearm motion. Finally we tested the wearable system and game on healthy subjects and stroke patients and analyzed the differences in subject performance which reflected their impairment level. We expect that our self-powered rehabilitation system could be used as a practical solution for home based elbow rehabilitation therapy.