High Rate Communication over One-Bit Quantized Channels via Deep Learning and LDPC Codes

Abstract

This paper proposes a method for designing error correction codes by combining a known coding scheme with an autoencoder. Specifically, we integrate an LDPC code with a trained autoencoder to develop an error correction code for intractable nonlinear channels. The LDPC encoder shrinks the input space of the autoencoder, which enables the autoencoder to learn more easily. The proposed error correction code shows promising results for one-bit quantization, a challenging case of a nonlinear channel. Specifically, our design gives a waterfall slope bit error rate even with high order modulation formats such as 16-QAM and 64-QAM despite one-bit quantization. This gain is theoretically grounded by proving that the trained autoencoder provides approximately Gaussian distributed data to the LDPC decoder even though the received signal has non-Gaussian statistics due to the one-bit quantization.

Design of Deep Learning-Based 1-bit Transceiver With Oversampling and Faster-Than-Nyquist (FTN) Signaling

“…Fig. 1 presents a comprehensive system model that includes existing implementations performing one-bit ADC and oversampling at the receiver and FTN signaling at the transmitter [2]- [13]. Let b ∈ B s denote information bits to be transmitted, which is then channel-encoded using code rate r = s/k whose output is given as c ∈ B k .…”

“…and M Tx (th) as design constraints. Subsequently, the following problem formulation comprehensively summarizes the practical design objectives of one-bit ADC as given in [1][2][3][4][5][6][7][8][9][10][11][12][13]:…”

“…The design approach in this problem formulation recognizes the challenge of meeting the many constraints and limits of one-bit ADC systems. It organizes the design efforts of various existing state-of-the-art schemes for one-bit quantization channel [1][2][3][4][5][6][7][8][9][10][11][12][13]. However, the non-explicit representation of BER with respect to SNR and other parameters makes it a non-trivial optimization problem.…”

A Review of Deep Learning in 5G Research: Channel Coding, Massive MIMO, Multiple Access, Resource Allocation, and Network Security

“…Since 5G communications systems have adopted polar codes, specifically LDPC codes, DL has been used to discover methods to blindly identify LDPC codes [23], reduce the decoding delay [25], [36], [37], [38], analyze the trade-off of LDPC codes for channel coding [24], develop error correction codes for nonlinear channels [39] and optimize the decoding algorithm to solve a non-convex minimization problem [40]. As expected, when comparing traditional decoding to DL-based decoding, DL has a greater reward.…”

“…• Large spectral efficiency gain • Higher-order modulation formats are operable for one-bit receivers [39] decompression process to recover the original CSI for singleuser and multi-users hoping to improve reconstruction quality and feedback accuracy of the CNN compressed structural characteristics of the massive MIMO channel. Both 2D and 3D were used but for separate cases; 2D was for the singleuser case and 3D was for the multi-user case.…”