An issue of realizing dynamic spectrum access in cognitive radio network is the design of an efficient channel assignment mechanism. In this work, a selective channel assignment method is proposed to divide the channels between the Primary Users (Pus) and Secondary Users (SUs). There is need to build two first fit assignment models in one single model to treat each of the PU and SUs separately. Also, there should be a flexibility to decide which part of the channel to be centered around for the allocation of PU and SUs, an attempt is made to design such a model. The performance of the model is tested on the network that has converters and also on network that does not have any converters installed. The percentage of PU calls over all the calls is also varied to determine the change in blocking probabilities. Extensive simulation results show that the proposed selective channel assignment method performs better with respect to performance parameters such as the blocking probabilities, throughput and the channel usage frequency, when compared with the existing channel assignment methods such as first fit assignment and uniformly distributed random assignment methods.