Reduced-Complexity Decision-Directed Pilot-Aided Tracking of Doubly-Selective Channels

We consider tracking a doubly-selective channel given a transmission stream which alternates between data and pilot blocks. Current and previous pilots, as well as previouslydecoded data blocks, are used for estimation of channel parameters. The optimal Kalman and Wiener estimators for this scenario are computationally expensive. In response, we propose a novel two-stage estimation technique whose performance is close to that of the optimal estimators but whose complexity is significantly less. The first stage finds smoothed estimates of the channel during previous frames, while the second stage uses these smoothed estimates for channel prediction within the current frame.