Nonlinear channel estimation in fading OFDM systems

We propose a (stochastic) nonlinear filtering approach to the channel estimation in OFDM communications affected by frequency-selective fading. The complex envelopes of a set of pilot tones are oversampled in the symbol interval and their phases and frequencies are tracked using a bank of nonlinear filters, each one coupled with a Kay frequency estimator. Symbol detection is performed by a bank of correlators and a decision block driven by the phase/frequency estimates. These estimates are updated in each symbol interval; thus, the receiver can withstand rapidly varying channels and track fast changes of the carrier frequency (due to Doppler shift and/or oscillators instabilities)