Subspace Projection-based OFDM Channel Estimation

In this paper, we investigate the benefits of pre-processing received data by projection on the performance of channel estimation for orthogonal frequency division multiplexing (OFDM) systems. Projecting data onto its signal subspace will reduce the additive noise energy in the data. Least square (LS) estimation is a low-complex algorithm for training-based OFDM systems and the lower bound on the mean-square error of it is proportional to the noise variance. So, after the received data is pre-processed (projected onto its signal subspace), LS channel estimation on the pre-processed data will increase the performance of channel estimation. This method can also work in multiple-input and multiple-output (MIMO) case. Performance analysis and simulation results show that the proposed algorithm has a considerably smaller complexity than the linear minimum mean square error estimation while having almost the same performance.