High mobility support of multiple-input multiple-output (MIMO)-orthogonal frequency division multiplexing (OFDM) systems plays a major role in wireless communication systems. Conventional detection approaches for MIMO-OFDM systems assume the channel is constant for at least one OFDM symbol duration. However, the time variation of the channel on account of mobility over an OFDM symbol period causes a loss of subchannel orthogonality which results in inter-carrier interference. In this paper, we propose the space alternating generalized expectation maximization (SAGE) based signal detection for MIMO-OFDM systems and compared by the zero forcing (ZF), minimum mean square error (MMSE), vertical Bell laboratories layered space-time (VBLAST) detection methods by considering complexity and performance tradeoffs. Simulation results show that the SAGE based detection has comparable performance with the VBLAST while needs lower computational complexity. It is also shown that both SAGE and VBLAST have a clear BER performance advantage over the MMSE and ZF detections. We also investigate the sensitivity of detection MIMO-OFDM signals to channel estimation errors by the application of a simple frequency domain based pilot based channel estimation. It is shown that the superiority of the SAGE scheme is also valid in the case of channel imperfections.