In order to optimize the multi-user subcarrier allocation of cognitive wireless network, it is converted into a constraint optimization problem. A chaotic immune optimization algorithm is proposed to solve it. The key techniques and implementation processes are given. The operators, such as coding, clonal, crossover, and mutation, are designed. The experimental results show that in conditions of user rate, the bit error rate and inference constraints, the algorithm minimizes the total transmit power and converges rapidly. It can obtain the better allocation scheme and improve the utilization efficiency of high frequency spectrum.