Abstract: The end to end (E2E) throughput can be enhanced through channel allocation in cognitive relay network. In this work, channel allocation for a three-node cognitive relay network was investigated A suboptimal allocation method was proposed when decode-and-forward (DF) was exploited by relay node. It ordered the channels in terms of channel gains，then successively allocated relay channels. The optimal and suboptimal methods were also proposed under amplify-and-forward (AF) protocol. The suboptimal method allocated relay channel by comparing E2E throughput when it was treated as S-R or R-D link of a dual-hop channel with that as a conventional relay channel. Both suboptimal methods decrease the computational complexity with minor throughput loss comparing with the optimal ones. Numerical results were provided to verified the validity of those methods by demonstrate and compare the performance of channel allocation algorithms under AF and DF relaying techniques. We also attempted to compare the computation complexities of optimal and suboptimal algorithms by observing simulation time. For a given channel quantity, ratios of allocated channels are given. It shows that relay channels are allocated as direct transmission channels with a ratio of 1/3. It is also observed that relay channels are rarely allocated as dual-hop channels under AF relaying technique.