TL-ANN Based Nonlinear Equalization for Multi-Users Radio Over Fiber System

To mitigate nonlinear distortions in multi-user analog radio over fiber (A-RoF) system, a nonlinear equalizer (NLE) which combines transfer learning (TL) and waveform regression-based artificial neural network (ANN) is proposed and validated experimentally in this research. Unlike traditional ANNs,...

Full description

Saved in:
Bibliographic Details
Published in:Journal of lightwave technology 2023-03, Vol.41 (5), p.1399-1405
Main Authors: Liu, Enji, Yu, Zhenming, Song, Yunyu, Sun, Kaixuan, Huang, Hongyu, Yin, Feifei, Zhou, Yue, Xu, Kun
Format: Article
Language:English
Subjects:
Artificial neural networks
Compatibility
Costs
Equalization
Equalizers
Nonlinear distortion
nonlinear equalization
Nonlinear optics
OFDM
Optical distortion
Optical receivers
Photonics
Radio
radio over fiber
Regression
Training
transfer learning
Waveforms
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:To mitigate nonlinear distortions in multi-user analog radio over fiber (A-RoF) system, a nonlinear equalizer (NLE) which combines transfer learning (TL) and waveform regression-based artificial neural network (ANN) is proposed and validated experimentally in this research. Unlike traditional ANNs, waveform regression ANNs can be regarded as waveform training ANNs and avoid complex-value training. In this paper, we first analyze the modulation distortions that appear in multi-user RoF systems and then propose a waveform regression ANN-based nonlinear equalizer (ANN-NLE). Subsequently, we use TL to reduce training costs and increase the compatibility of the ANN-NLE. Finally, an A-RoF system with multicore fiber is adopted to verify the proposed approach. According to the simulation and experimental results, the bit error ratio (BER) of a multi-user signal is decreased to the threshold of 3.8 × 10 −3 , and the improved optical receiver sensitivity is greater than 1.5 dB. This proposed method could realize nonlinear equalization and display good compatibility with different systems.
ISSN:0733-8724
1558-2213
DOI:10.1109/JLT.2022.3221166