Abstract: Due to the fact that in traditional relay gain allocation (RGA) schemes only the partial system parameters are included, the energy efficiency of cognitive relay systems is very low. To improve the performance of amplify-and-forward (AF) relay cooperation cognitive radio (AF-R-CR), a novel RGA scheme is proposed firstly, which is based on the capacity constraint of primary systems and the average symbol error ratio (SER) minimization of secondary systems. Then, for the proposed RGA scheme, we obtain the exact optimization relay gain (amplifier factor) as well as the corresponding lower and upper bounds that are given in closed-form expressions. Finally, based on the achieved optimization RGA scheme, the average SER of the considered AF-R-CR systems is derived. The derivations show clearly that in the proposed RGA the effect of the secondary source-relay links and the transmit power of the secondary source is considered, but also the one of the secondary relay-destination links and the parameters of primary systems. This yields the system resources exploited in an effective means, and the system performance improved greatly.