Abstract: An iterative channel estimation and detection scheme is proposed for single carrier uncoordinated multiuser/multi-input multi-output (MIMO) system using superimposed training (ST). At first, the multiple antennas mutual interferences are eliminated by utilizing training sequences with all-zero cross-correlation property, which are designed based on transform domain approach. The channel estimation is performed based on first-order statistics of the observations in time domain. Then a joint iterative channel estimation and symbol detection method is addressed, where the detected sequences are used as additional “training sequence” to reduce the self-interferences from information sequences. Compared to the existing ST schemes, the new proposed scheme has more flexible training sequences, better channel mean square error (NCMSE) and BER performance at low SNR, but retains a lower computation complexity. Simulation results corroborate the effectiveness of the new scheme.