In this paper,we study cross-layer scheduling scheme on multimedia application which considers both streaming traffic and data traffic over cognitive ad hoc networks.A cross-layer design is proposed to optimize SU’s utility,which is used as an approach to balance the transmission efficiency and heterogeneous traffic in cognitive ad hoc networks.A framework is provided for utility-based optimal subcarrier assignment,power allocation strategy and corresponding modulation scheme,subject to the interference threshold to primary user(PU) and total transmit power constraint.Bayesian learning is adopted in subcarrier allocation strategy to avoid collision and alleviate the burden of information exchange on limited common control channel(CCC).In addition,the M/G/l queuing model is also introduced to analyze the expected delay of streaming traffic.Numerical results are given to demonstrate that the proposed scheme significantly reduces the blocking probability and outperforms the mentioned single-channel dynamic resource scheduling by almost 8%in term of system utility. In this paper, we study cross-layer scheduling scheme on multimedia application which determines both streaming traffic and data traffic over cognitive ad hoc networks. A cross-layer design is proposed to optimize SU’s utility, which is used as an approach to balance the transmission efficiency and heterogeneous traffic in cognitive ad hoc networks. A framework is provided for utility-based optimal subcarrier assignment, power allocation strategy and corresponding modulation scheme, subject to the interference threshold to primary user (PU) and total transmit power constraint. Bayesian learning is adopted in subcarrier allocation strategy to avoid collision and alleviate the burden of information exchange on limited common control channel (CCC) .In addition, the M / G / l queuing model is also introduced to analyze the expected delay of streaming traffic. given to demonstrate that the proposed scheme significantly reduces the blocking probability and outperforms the mentioned sing le-channel dynamic resource scheduling by almost 8% in term of system utility.