The application of the multiuser multiple-input multiple-output (MIMO) technique with orthogonal frequency division multiplexing (OFDM) signal transmission to future wireless communications systems has attracted much attention. In the uplink, the difference in the arrival timings among the signals from multiple wireless terminals (WTs) causes significant transmission quality degradation. In this paper, we propose a novel multiuser detection (MUD) technique that overcomes this problem. The proposed method discards the samples at both ends of each fast Fourier transform (FFT) window. In addition, it overlaps the FFT windows to recover all consecutive samples. Thus, it enables a signal separation regardless of the difference in the arrival times among WTs with an acceptable level of calculation complexity. The achievable bit error rate (BER) performance of the proposed MUD method is evaluated by computer simulation in a frequency selective fading channel. The results show that compared to the conventional method the proposed MUD method reduces the required Eb/N0 by 2.6 dB to achieve the BER of 10-4, when 16QAM and rate 2/3 convolutional codes are used.