Low Complexity Two-Stage Soft/Hard Decoders

TL;DR

This paper introduces low complexity two-stage soft/hard decoders using ADBP for wireless systems, achieving near-optimal performance with reduced complexity, especially in challenging channel conditions.

Contribution

It proposes novel low complexity decoding schemes based on ADBP applied to two-stage encoding, addressing limitations of current BICM techniques in wireless communications.

Findings

Negligible performance loss with hard decoding for most bits in PAM^2 over AWGN.

Number of soft bits needed increases with Rician factor, up to three in Rayleigh channels.

ADBP reduces detection and decoding complexity effectively.

Abstract

Next generation wireless systems will need higher spectral efficiency as the expected traffic volumes per unit bandwidth and dimension will inevitably grow. As a consequence, it is necessary to design coding schemes with performances close to the theoretical limits, having high flexibility and low complexity requirements at transmitter and receiver. In this paper, we point out some of the limitations of the Bit Interleaved Code Modulation (BICM) technique which is the state of the art adopted in several standards and then propose some new lower complexity alternatives. These low complexity alternatives are obtained by applying the recently introduced Analog Digital Belief Propagation (ADBP) algorithm to a two stage encoding scheme embedding a hard decoding stage. First we show that for PAM$^2$ type constellations over the AWGN channel, the performance loss caused by using a hard decoded stage for all modulation bits except the two least protected is negligible. Next, we consider the application of two stage decoders to more challenging Rician channels, showing that in this case the number of bits needed to be soft decoded depends on the Rician factor and increases to a maximum of three bits per dimension for the Rayleigh channel. Finally, we apply the ADBP algorithm to further reduce the detection and decoding complexity.