Fractional Interference Alignment: An Interference Alignment Scheme for Finite Alphabet Signals

TL;DR

This paper introduces Fractional Interference Alignment (FIA), a novel scheme optimized for finite alphabet signals that improves error rate performance over traditional IA by using a non-linear detection approach and a new metric called SpAC.

Contribution

The paper proposes FIA, a new interference alignment scheme tailored for finite alphabet signals, utilizing non-linear detection and a novel performance metric, SpAC.

Findings

FIA achieves better error rates than traditional IA with Gaussian assumptions.

FIA allows flexible SpAC values between 0 and 1, including the conventional 1/2.

Numerical results show significant performance improvements in bit error rate.

Abstract

Interference Alignment (IA) is a transmission scheme which achieves 1/2 Degrees-of-Freedom (DoF) per transmit-antenna per user. The constraints imposed on the scheme are based on the linear receiver since conventional IA assumes Gaussian signaling. However, when the transmitters employ Finite Alphabet (FA) signaling, neither the conventional IA precoders nor the linear receiver are optimal structures. Therefore, a novel Fractional Interference Alignment (FIA) scheme is introduced when FA signals are used, where the alignment constraints are now based on the non-linear, minimum distance (MD) detector. Since DoF is defined only as signal-to-noise ratio tends to infinity, we introduce a new metric called SpAC (number of Symbols transmitted-per-transmit Antenna-per-Channel use) for analyzing the FIA scheme. The maximum SpAC is one, and the FIA achieves any value of SpAC in the range [0,1]. The key motivation for this work is that numerical simulations with FA signals and MD detector for fixed SpAC (=1/2, as in IA) over a set of optimization problems, like minimizing bit error rate or maximizing the mutual information, achieves a significantly better error rate performance when compared to the existing algorithms that minimize mean square error or maximize signal-to-interference plus noise ratio.