Near LMMSE channel estimation performance with artificial channel at receiver for OFDM systems

This paper proposes a LMMSE-based channel estimator which, unlike the classical LMMSE estimator, does not require the covariance matrix of the channel nor its estimation. Actually, we add at the receiver side a fully adjustable filter, which acts like an artificial channel and hides the physical channel. We then perform an LMMSE estimation of the sum of the physical and artificial channel using the filter covariance matrix and the channel estimation is obtained by subtracting the filter. Theoretical developments, shown in this paper, prove that the performance of the proposed solution can be driven by the parameters of this additional filter and can reach the one of the theoretical LMMSE estimator closely. Simulations, proposed in a DRM context, also display the validity of this technique for both preamble-based and scattered pilot distributions. It is shown that the proposed solution is only about 1 dB from the LMMSE technique in term of Minimum Mean Square Error (MMSE) and Bit Error Rate (BER).