A reduced-rank space-time channel estimation scheme is presented for
uplink multiuser DS-CDMA (direct-sequence code-division multiple access) systems in dispersive MIMO (multiple input and multiple output) channels. The proposed method was based on the AP method with training sequences. It is shown that both the mean squared error of channel estimation and the bit error probability can be improved via rank order selection. The optimal performance can be achieved using the generalized likelihood ratio test (GLRT).