In this paper, an adaptive optimization scheme for DCF (distributed coordination function) is proposed to enhance the network performance. The scheme is based on the channel sensing result for network state information and thus it is named CSB (channel sensing backoff). The key idea to approach optimal performance dynamically in the proposed scheme is that the transmission attempt from the DCF is filtered by an adjustable probability P_T, which is dynamically adapted to reflect the current channel competing level among the network stations. Unlike other proposals for the DCF optimization, CSB does not need to perform complex on-line estimation of the number of active stations in the network, and can make adaptive tuning always toward a certain optimization object under various network states. Detailed performance evaluation of the scheme is carried out through NS-2 simulation. Simulation results show that the scheme can effectively adapt to both network size and packet length changes in the network, and simultaneously achieve performance improvements in several aspects including system throughput, collision probability, delay, delay jitter, fairness and so on.