Iterative Frequency-Domain Joint Channel Estimation and Data Detection of Faster-Than-Nyquist Signaling

In this paper, we propose semi-blind iterative frequency domain joint channel estimation (CE) and data detection (DD) of faster-than-Nyquist signaling (FTNS). The proposed scheme achieves low-complexity operation, while maintaining a performance close to that of the perfect channel state information scenario. More specifically, we derive low-complexity frequency-domain CE for the faster-than-Nyquist pilot (FTNP) transmission scenario, where the symbol duration of a pilot block is lower than the Nyquist-criterion-based pilot transmission. Moreover, we propose a serially concatenated channel-encoded FTNS transceiver that takes into account FTNS-specific colored noise effects. The proposed low-complexity receiver carries out soft-decision frequency-domain equalization with the aid of the minimum-mean square error criterion while whitening the colored noise. As explicit benefits of the proposed frequency-domain CE for the FTNP, the demodulated FTNS data block can also be used as a long pilot block, so the iterative joint CE and DD becomes realistic. Simulation results demonstrate that the proposed two-stage-concatenated FTNS system relying on binary phase-shift keying-based FTNP and FTNS-data symbols achieves a better error-ratio performance than previous systems that do not consider colored noise effects in the high-symbol-packing FTNS regime.