This paper applies TCP/IP behavior characteristics and wireless link BER change to establish a wireless heterogeneous model with binary-state Markov error model, to describe the practical congestion degree in network more accurately. Then, the price mechanism of REM algorithm is adopted to design a congestion control method for rapid response and queue cache rate. By comparing the receiving rate of access point, link bandwidth and the rate of queue on node buffer are used to control queue length, so the adaptability of queue changes is improved. Then, by means of analyzing lacking price production mechanism in standard random exponential marking algorithm, the maximum frame is introduced into REM price mechanism so that it can be more adapted to wireless environment. In addition, queue cache ratio and link virtual capacity are taken as congestion measurement index and a new price-based congestion control strategy is proposed. Simulation results show that our scheme shows better performance and advantage on adaptability of dynamic bandwidth as well as QoS assurance of multimedia transmission delay, which also effectively improves the utilization of link bandwidth.