OFDM CIR Estimation with Unknown Length via Bayesian Model Selection and Averaging

This work presents new CIR estimators for OFDM systems over frequency selective channels, where the length of the CIR is unknown a-priori. We derive the MMSE estimator for this model using two different criteria, namely, Bayesian model averaging and Bayesian model order selection. Estimation of the CIR length enables adapting the CP length to the changing propagation environment, resulting in increased throughput due to shorter CP. Simulation results under different channel conditions demonstrate the robustness of the estimators.