Abstract: The performance of a multi-input multi-output orthogonal frequency division multiplexing (MIMO-OFDM) system can be improved when the channel state information (CSI) is available for beamforming at the transmitter. However, even with the quantization of the CSI, the feedback requirement is still prohibitively unaffordable in a frequency division duplexing based MIMO-OFDM system with a large number of subcarri- ers. In this paper, two schemes with limited feedback, i.e., a nonlinear interpolation and a modified clustering based transmit beamforming schemes, are proposed. In the first method, the beamforming vectors for the non-pilot subcarriers in each cluster are constructed via that of the pilot subcarriers through nonlinear interpolation. In the second method, the beamforming vector for each cluster is selected to optimize the performance of a group of subcarriers which are chosen from that cluster based on equal spacing or a predefined random matrix. Simulations demonstrate that the proposed nonlinear interpolation and the modified clustering based transmit beamforming schemes outperform the existing limited feedback methods with the same feedback.