The demand of quality of service (QoS) for multimedia transmission over wireless network raises huge challenges such as time-varying channel conditions, limited resources, tight delay constraints, high bandwidth demand and complex protocol design. Layered design is proven to perform well in wired networks. However, doing so is not suitable in wireless networks due to the dynamics in wireless medium. On top of that, sensing activity should also be considered explicitly in cognitive UWB network to avoid interference with the primary users. In this paper, a new cross layer design (CLD) approach is proposed to achieve the goal of improving the received video quality while minimizing the packet loss. The proposed CLD consists of sensing module, adaptive resource allocation module and adaptive quantization scale module. Simulation results showed that the proposed CLD reduced the job failure rate (JFR) significantly when compared to the non-CLD approach, basic CLD, PRR based resource allocation (PRR-RA) with queue insensitive and PRR-RA with queue sensitive-fixed Q-value.