The lack of wireless spectrum resource is one of the main factors constraining the further development of wireless communication, and this has been receiving increasing research attention. Cognitive radio (CR) has been demonstrated as a good solution to improve resource efficiency. Moreover, orthogonal frequency division multiplexing (OFDM) technology has become the preferred technique in CR systems due to its flexible dynamic resource allocation. We propose a distributed resource allocation framework based on the particle swarm optimization (PSO) algorithm for CR networks utilizing an OFDM-based system. Our main aim is to maximize the system throughput subject to constraints and put forward a computationally efficient, suboptimal power allocation algorithm for multi-user OFDM CR system. Simulation results are provided to evaluate the performance of the PSO algorithm in comparison with the optimal Lagrange dual method.