OFDM With Iterative Blind Channel Estimation

Abstract

A new blind channel estimation technique is presented for uncoded orthogonal frequency-division multiplexing (OFDM) systems. Instead of using pilots to sound the channel, a decision algorithm first makes primary estimates of the data symbol for each subcarrier based on a constrained linear minimum mean square error (MMSE) criterion. Then, these estimates are applied to optimal MMSE channel estimation. The technique requires only one value from the time–frequency correlation of the channel transfer function. Performance is evaluated by simulation so that comparison can be made with known optimal coherent/differential detection. Compared with known decision-directed Kalman-based estimation and two pilot-aided OFDM schemes (block pilots and comb pilots), the presented technique performs better for regions with mid to high signal-to-noise ratios (SNRs). Its robustness to the time variation of the channel is also quantified by simulation, showing only small degradation in performance relative to the quasistatic case of wireless local area network (WLAN) systems. Finally, the impact of covariance assumptions in the channel modeling is quantified using simulation, offering a feel for the performance with mismatch between the channel model and the receiver assumptions.