Functional rehabilitation aims at recovering the locomotion dysfunction of the human body by the physical therapy. The objective of this paper was to develop interactive and connected rehabilitation systems as a system of systems for monitoring the bio-feedbacks of the human musculoskeletal system during the rehabilitation exercises. Video-based non-contact system as Kinect sensor was used to get kinematics data of the human body. Generic and subject-specific avatar representations were integrated. Rehabilitation exercises will be designed as serious games to motivate the end users. Our first prototype was focused on the rehabilitation exercises of the lower limb. Software development and experimental aspects of our proposed solution were presented and discussed. Our system would be of great interest in the supervision of physical therapy exercises in clinical as well as in non-clinical environments (e.g. rehabilitation at home). As perspectives, multi-sensor fusion between Kinect sensor and other kinematics-based sensors like Shimmer ones will be investigated to get an accurate 3D joint motion. Electromyography (EMG) signals will be also used to monitor the muscle functions. Moreover, specific device will also be developed to facilitate the sensors set up and motion monitoring.