Deep-learning-based Signal Detection for Channel Coded Faster-than-Nyquist Transmission.

Faster-than-Nyquist (FTN) is a promising paradigm to improve bandwidth utilization at the expense of additional intersymbol interference (ISI). In this letter, we develop a signal detection architecture based on deep learning (DL) for FTN system, which employs the sliding window and works without any iteration. Furthermore, to better fit the channel coded scenarios, we develop a signal reconstruction method based on hard decision and FTN mapping. To the best of our knowledge, this is the first attempt to apply DL to channel coded FTN signal detection. As demonstrated by simulation results, in the uncoded scenario, our proposed DL-based detection can achieve a near-optimal bit error rate (BER) performance and shows the potential in high order modulations. Also, the proposed detection can achieve great performance gains from start-of-the-art channel coding scheme. Finally, the proposed DL-based detection has the robustness to signal to noise ratio (SNR). In a nutshell, DL has been proved to be a powerful tool for FTN signal detection.