Deep Learning Assisted Sum-Product Detection Algorithm for Faster-than-Nyquist Signaling

Abstract

A deep learning assisted sum-product detection algorithm (DL-SPA) for faster-than-Nyquist (FTN) signaling is proposed in this paper. The proposed detection algorithm concatenates a neural network to the variable nodes of the conventional factor graph of the FTN system to help the detector converge to the a posterior probabilities based on the received sequence. More specifically, the neural network performs as a function node in the modified factor graph to deal with the residual intersymbol interference (ISI) that is not modeled by the conventional detector with a limited number of ISI taps. We modify the updating rule in the conventional sum-product algorithm so that the neural network assisted detector can be complemented to a Turbo equalization. Furthermore, a simplified convolutional neural network is employed as the neural network function node to enhance the detector's performance and the neural network needs a small number of batches to be trained. Simulation results have shown that the proposed DL-SPA achieves a performance gain up to 2.5 dB with the same bit error rate compared to the conventional sum-product detection algorithm under the same ISI responses.

Time-Domain vs. Frequency-Domain Equalization for FTN Signaling

“…Conventionally, (6) can be calculated by employing classic trellis-based algorithms having a complexity exponentially increasing with L. According to (6), the metric of the erroneous sequence (path) x + e is then given by…”

“…Recent advances of FTN signaling include the signal detection, code design, and related applications. For instance, in [6], a deep learning assisted FTN detector was presented, where a deep neural network (DNN) is concatenated to the factor graph of FTN signaling. By applying a revised message updating rule, this DNN-based detector is able to obtain superior error performance compared to that of conventional FTN detectors with almost the same detection complexity.…”

A Novel Sum-Product Detection Algorithm for Faster-than-Nyquist Signaling: A Deep Learning Approach

“…It can be seen that compared to the training method in Section III.B, the compatible training technique includes two more stages for processing. This indicates that the input training sequences {Ψ (1) + 0, Ψ (2) + 0, ...} in the previous sections are now changed to {Ψ (1) + Υ (1,1) , ..., Ψ (1) + Υ (1,|Ω|) , Ψ (2) + Υ (2,1) , ...}. These two stages are the preprocessing of the NN training.…”

“…Recently, deep learning supplemented communication systems have shown the potential to further enhance the system's performance [1], [20]- [23]. In particular, for the detection and decoding algorithms, the research on autoencoders [24] and the NN optimization schemes which transform the FGs into NN systems [25], [26], has drawn significant interests.…”

Design and Analysis of Multi-User Faster-Than-Nyquist-DCSK Communication Systems over Multi-Path Fading Channels

“…In [ 23 ], output-retainable convolutional codes (ORCCs) were used for channel memorization to reduce the decoding complexity. A new approach based on a deep learning-assisted sum-product detection algorithm (DL-SPA) was presented in [ 24 ] for faster detection convergence.…”