State-space estimation of rapidly-varying carrier frequency offsets in OFDM systems

The problem with the application of parametric algorithms to estimate rapidly-varying carrier frequency offsets (CFO) in OFDM systems is that they are rather bandwidth-inefficient as a substantial part of the received signal is occupied with pilot symbols. Herein, we propose to minimize this drawback, i.e., we aim to increase the spacing between pilot symbols by using a truncated Taylor's series to predict the CFO, where the derivatives up to order d-1 of the series terms are recursively estimated with a d-order Kalman-Bucy filter (KBF). For large prediction intervals simulations show that the bit error floors can be significantly reduced when the conventional linear non-recursive CFO predictor is replaced with a fourth-order KBF.