可见光通信系统失真信号的无监督还原方法

Unsupervised recovery method of distorted signals for visible light communication

发光二级管的非线性特性、环境光干扰以及信道噪声等因素会引发可见光通信系统出现非线性失真的现象,导致可见光通信系统的误码率升高。针对该问题,提出一种新的深度神经网络模型,并基于该模型对可见光通信系统的非线性失真符号进行无监督地还原处理。建立了可见光通信系统的数学模型,分析了引起信号出现非线性失真现象的多种因素;利用循环神经网络学习可见光信号序列的短期相关性,利用门控循环单元学习可见光信号序列的长期相关性;采用密集网络学习输入光信号与接收光信号之间的非线性映射关系。仿真结果表明,所提神经网络能有效降低可见光通信系统的误码率,并且在不同调制阶数与不同带宽条件下均实现了较好的效果。

The factors of non-linear character of LED,ambient light and channel noise lead to non-linear distortion in visible light communication system,and then the bit error rate of the visible light communication system increases.Aiming at this problem,a new deep neural network model is proposed,and the non-linear distorted symbols of the visible light communication system are unsupervised recovered based on this model.First of all,the mathematical model of the visible light communication system is constructed,and multiple factors of non-linear distortion are analyzed;then,the recurrent neural network is utilized to learn the short term correlation of the visible light signal sequence,and the gated recurrent unit is utilized to learn the long term correlation of the visible light signal sequence;Finally,a dense network is adopted to learn the non-linear mapping relationship between the input visible light signals and the received visible light signals.Simulation results show that the proposed neural network can reduce the bit error rate of the visible light communication system effectively;it also achieves better effect on different modulation orders and different bandwidths.