Semi-blind equalization for OFDM systems over fast fading channels

We consider semi-blind data detectors for OFDM systems over fast fading channels. Three maximum-likelihood (ML) data detectors (MLD) are derived assuming exact channel correlation. The first two exploit the correlation among pre-DFT and post-DFT received signals and the third one average the ML metric over channel impulse response (CIR). We show that the first two MLDs for input data symbols are given by an integer least-squares (LS) minimization problem, which can be efficiently solved by V-BLAST detection (suboptimal) or by a sphere decoder (SD) (optimal). We contribute a low-complexity near-ML SD and an iterative detector for the third MLD. Despite its low complexity, MLDl performs robustly for normalized Doppler rates less than 3%, which meets the requirement of many practical systems. With iterative detection, both MLD1 and MLD3 can exploit time-diversity.