Biomedical connected objects like kinematic sensors have been commonly used for patient monitoring in many clinical applications. Moreover, serious games have become widely used to improve patients‘ motivation during functional rehabilitation. In this work, we developed and evaluated a new engineering system as a solution for functional rehabilitation at home. A multi-sensor fusion between Kinect camera and inertial sensors was developed to animate a 3D avatar during rehabilitation and to estimate kinematic data of different joints for clinical monitoring. Two serious game scenarios were designed for upper and lower limb rehabilitation. The developed system was evaluated through patient kinematic data and a questionnaire-based approach with a panel of eight post-stroke patients and four clinical experts. The evaluation of the system showed that multi-sensor fusion provides useful data for clinical follow-up. The virtual game scenarios lead to a high level of immersion for patients. Feedbacks from clinical experts concerning the system's GUIs and the clinical relevance of the acquired data for each rehabilitation session are positive. The developed system paves the way to deploy recent technologies, such as multi-sensor fusion and serious games, as a solution for home-based rehabilitation, which can optimize the benefit of the involved patients and medical experts.