EM-Based ML Estimation of Fast Time-Varying Multipath Channels for SIMO OFDM Systems

This paper investigates the problem of fast time-varying multipath channel estimation over single-input multiple-output orthogonal frequency-division multiplexing (SIMO OFDM)-type transmissions. We do so by tracking the variations of each complex gain coefficient using a polynomial-in-time expansion. To that end, we derive the log-likelihood function (LLF) in both the data-aided (DA) and non-data-aided (NDA) case. The DA ML estimates are found in closed-form expressions and then used to initialize the expectation maximization (EM) algorithm that is used to iteratively maximize the LLF in the NDA case. We also introduce an alternative initialization procedure that requires less pilot symbols as compared to the DA ML-based solution without incurring a significant performance loss. Simulation results show that the proposed EM-based estimator converges within few iterations providing accurate estimates for all multipath gains, thereby resulting in significant BER gain as compared to the DA least square (LS) technique.