Particle swarm optimization algorithm plays an important role in cognitive radio spectrum allocation. It can only work in continuous and unconstrained condition. And there are the problems of premature convergence and slow convergence speed, which are not effectively resolved. In order to effectively solve these problems, this article analyzes and improves the premature convergence problem of particle swarm optimization (PSO), successfully applying particle swarm optimization (PSO) algorithm to solve the problem of spectrum allocation. Considering the system total broadband returns and the fairness of user access, the corresponding objective function is established, the feasibility and superiority of this algorithm are proved.