基于神经网络的后均衡方法在水下可见光通信中的应用

Application of neural network based post equalization methods in underwater visible light communication

随着人类进一步推进对海洋的探索,人们对水下通信技术提出更高的要求。水下可见光通信(UVLC)具有高速、低延迟和高保密性的优点,吸引了广大研究者的关注。然而,在水下可见光通信中,水下信道环境很复杂,湍流、漫射、散射等因素严重影响信号传输的质量,同时,由器件引起的非线性失真限制着系统的性能。神经网络能够拟合复杂的非线性问题,开始被应用于可见光通信的信号均衡中。本文介绍了基于神经网络的后均衡方法在水下可见光通信中的应用,对比分析了DNN、DBMLP和TFDNet三种网络在非线性均衡方面的性能,并且讨论了不同计算复杂度对三者的影响,为在实际应用中采用不同的均衡方法提供参考。

Great demands have been placed on underwater communication technology as human mankind keep on exploring the ocean.With the advantages of high speed,low latency and confidentiality,underwater visible light communication(UVLC)has attracted more and more attention.However,the underwater environments are complex,and scattering,turbulence and diffusion will affect the quality of the signal.Besides,nonlinear distortion caused by the system devices deteriorates the system performance of UVLC.As neural network has ability to solve complex nonlinear problem,it is employed to finish signal equalization in visible light communication.This paper introduces the application of neural network-based post equalization methods in UVLC,and then analyzes the performance of DNN,DBMLP and TFDNet in terms of nonlinear equalization.The influence of different computational complexity on these methods are also discussed.We wish this paper could provide some reference for selecting different equalization methods in practical application.