Generalized Nearest Neighbor Decoding: General Input Constellation and a Case Study of Interference Suppression

TL;DR

This paper introduces a generalized nearest neighbor decoding (GNND) approach for channels with arbitrary input constellations, demonstrating its effectiveness in multiuser interference suppression and near-optimal information rates.

Contribution

It derives optimal GNND criteria for general input constellations and develops a practical coded modulation scheme that outperforms traditional linearization methods.

Findings

GNND achieves near channel mutual information rates in multiuser uplink.

The proposed scheme outperforms channel linearization-based methods.

Optimal GNND criteria are expressed via conditional moments matching.

Abstract

In this work, generalized nearest neighbor decoding (GNND), a recently proposed receiver architecture, is studied for channels under general input constellations, and multiuser uplink interference suppression is employed as a case study for demonstrating its potential. In essence, GNND generalizes the well-known nearest neighbor decoding, by introducing a symbol-level memoryless processing step, which can be rendered seamlessly compatible with Gaussian channel-based decoders. First, criteria of the optimal GNND are derived for general input constellations, expressed in the form of conditional moments matching, thereby generalizing the prior work which has been confined to Gaussian input. Then, the optimal GNND is applied to the use case of multiuser uplink, for which the optimal GNND is shown to be capable of achieving information rates nearly identical to the channel mutual information. By contrast, the commonly used channel linearization (CL) approach incurs a noticeable rate loss. A coded modulation scheme is subsequently developed, aiming at implementing GNND using off-the-shelf channel codes, without requiring iterative message passing between demodulator and decoder. Through numerical experiments it is validated that the developed scheme significantly outperforms the CL-based scheme.