The idea of cognitive network enables a distributed system to observe, learn, and decide at the system-level scope, thus achieves better performance than traditional distributed systems that operate at the local scope. In this paper we would like to enable the concept of cognitive network in an UPnP-based wireless home network environment. We found that the popular UPnP QoS architecture possesses some features of the cognitive network, while it is deficient in the learning capability. Consequently, we refine the UPnP QoS architecture with a cognitive module that utilizes the network topology and spatiotemporal traffic statistics to determine an end-to-end traffic for multimedia applications. Based on the idea, we propose a two-phase failover mechanism for wireless links in the UPnP-based wireless home network environment. We believe that our work can contribute to the development and deployment of QoS-required multimedia applications in a wireless smart home environment.