Butterfly Neural Filter Applied to Beamforming

Tiago F. B. de Sousa,Marcelo A. C. Fernandes

doi:10.1109/access.2019.2929590

Abstract

The butterfly neural beamformer (NB-Butterfly) is a new adaptive multiple-antenna spatial neural filter inspired on the neural butterfly equalizer (NE-Butterfly), a filter intended to equalize any channel that has real or complex taps, whether linear or nonlinear. Due to the broad use cases of the NE-Butterfly, the objective in this paper is to introduce this novel beamforming filter, the NB-Butterfly and analyze its performance by comparing to other neural and linear beamformers, while also presenting an enhanced training strategy that wasn’t present in the butterfly neural architecture before, which is called butterfly neural beamformer with joint error (NB-Butterfly-JE). The proposals are evaluated and compared for different types of channels in order to validate their performance in different use cases.

Highlights

Spatial filters with adaptive multiple-antenna processing are known to dramatically enhance wireless communication systems, being able to identify signals transmitted on the same carrier frequency that are sufficiently separated in the spatial domain [1], [2]
These algorithms use the minimum mean square error (MMSE) principle to achieve the desired output through the use of a known reference signal and the output of the filter, when considering space-division multiple access (SDMA), where channel configuration and spatial separation in angles of arrival (AOA) of the signal and other sources dictates the performance of the communication, what happens when the conditions aren’t favorable is that the individual sources become linearly inseparable [4]–[7]
The second is to compare its performance with other proposals, where a least mean square (LMS) beamformer and a beamforming adapted Bi-dimensional Neural Beamformer with Joint Error (BNB-JE) were used to have both a linear and a neural beamformer to serve as a comparison basis

Summary

INTRODUCTION

Spatial filters with adaptive multiple-antenna processing are known to dramatically enhance wireless communication systems, being able to identify signals transmitted on the same carrier frequency that are sufficiently separated in the spatial domain [1], [2]. The technique most commonly used in the processing of adaptive antennas is known as beamforming, where the general algorithm creates a linear combination of the received signals at different elements in an antenna array These algorithms use the minimum mean square error (MMSE) principle to achieve the desired output through the use of a known reference signal and the output of the filter, when considering space-division multiple access (SDMA), where channel configuration and spatial separation in angles of arrival (AOA) of the signal and other sources dictates the performance of the communication, what happens when the conditions aren’t favorable is that the individual sources become linearly inseparable [4]–[7]. The digital communication system where the beamformers were tested uses 4-QAM (quadrature amplitude modulation), which was simulated for four different channels that have varying levels of multi pathing and AOA between the source signal and its reflections, leading to linear and nonlinear configurations

COMMUNICATION SYSTEM MODEL

ARCHITECTURE

SIMULATIONS AND RESULTS

CONCLUSIONS