Dynamic spectrum allocation (DSA) in cognitive radio system has become a hot topic to solve the spectrum scarcity. Particle swarm optimization (PSO) has shown great flexibility in solving the spectrum allocation problem, and many related studies have been proposed. However, there is still room for improvement in the problem that PSO will have a slow convergence rate in the DSA and the latter particle search is easy to fall into the local optimum. In this paper, a dynamic time-varying spectrum allocation scheme based on chaotic binary particle swarm optimization is proposed. The improved algorithm introduces the idea of chaotic map to optimize the initial population and the optimal position of each generation of particles, and utilizes the global ergodicity of chaotic map to overcome the shortcomings of the algorithm. Based on the graph theory model, a mathematical model of dimensionality reduction spectrum allocation is built to realize DSA for users. The experimental results show that the convergence time cost of the improved algorithm is low, and the higher network rewards can be obtained under different experiments, and the fast and efficient spectrum allocation is realized to satisfy the demands of cognitive radio communication.