In this paper, we present a novel user grouping method for grouped OFDM-IDMA systems. Aiming at maximizing the system capacity, we adaptively distribute the users among the pre-allocated subcarrier groups according to their respective channel conditions. We analyze the achievable capacity of the system and formulate the optimization problem as a weighted bipartite matching problem. Kuhn-Munkres method is employed to solve the problem. The performance of the proposed scheme is evaluated through both theoretical analysis and simulation. It shows that with our proposed algorithm, the system throughput is greatly improved and is very close to the theoretical upper bound.