LDPC Code Design for Binary-Input Binary-Output Z Interference Channels

Abstract

In this paper, we explore code optimization for two-user discrete memoryless interference channels (DMICs) wherein the inputs and outputs of the channel are from a finite alphabet. For encoding, we employ irregular low-density parity-check (LDPC) codes combined with non-linear trellis codes (NLTCs) to satisfy the desired distribution of zeros and ones in the transmitted codewords. At the receiver sides, we adopt BCJR algorithm based decoders to compute the symbol-by symbol log-likelihood ratios (LLRs) of LDPC coded bits to be fed to message passing decoders. As a specific example, we consider the binary-input binary-output Z interference channel (BIBO ZIC) for which the transmitted and received signals are binary and one of the receivers is interference free. For a specific example of a BIBO ZIC, we examine the Han-Kobayashi inner bound on the achievable rate pairs and show that with a simple scheme of sending the messages as private one can achieve the sum-capacity of the channel. We also perform code optimization and demonstrate that the jointly optimized codes outperform the optimal single user codes with time sharing.