As one of common diseases among elderly, stroke often leads to motor dysfunction and even permanent disability. In order to improve the motor function of paralyzed extremity, patients can perform effective exercise for rehabilitation. This project aims to design a hybrid BCI (Brain-Computer Interface)–VR (Virtual Reality) system, including the motor imagery training program and the physical training program, to improve the effectiveness of post-stroke rehabilitation. The established system has been tested with high classification accuracy of motor imagery based on neural network, and high level of presence in user experience using questionnaires.