DiffServ (differentiated service) methods are always used at the edge of network to guarantee the QoS (quality of service) . Current label system of DiffServ is static and not flexible, it is a distinct drawback. In order to acquire intelligence and improve the Subjective QoS, knowledge based video quality guarantee model is provided and a prototype system is built. Different from others, our system is a knowledge-centered one. In our prototype system the video adaptation is guided intelligently with the knowledge acquired at different phases of knowledge creation process and (m,k) model is applied in label design so as to mapping to knowledge . Testing results have shown this system can satisfy the user preference and ensure video quality effectively. So it is proved that knowledge play an important role and bring more intelligence to our system.