In order to improve the downlink capacity of a multiuser multiple input multiple output system exploiting spatial multiplexing, this paper proposes two grouped multiuser diversity schemes. Aiming at the suppression of co-channel interference, both schemes select a group of users to share the same time-frequency channel unit in two steps. The first scheme iteratively selects the users based on a traditional multiuser scheduler - the PFS method. The second scheme first groups the users according to the cross-correlation of their channel response matrixes, then selects the group that maximizes the system capacity. Simulation results show that both schemes can provide fair scheduling and enhance the capacity performance. Moreover, the second scheme achieves higher capacity with more computational complexity comparing with the first one.