Research On Nonlinear Distortion Compensation Technology Of Visible Light Communication System

As a new communication method with green,safe and abundant spectrum resources,visible light communication(VLC)can efficiently relief the crisis of spectrum resource scarcity in RF communication.However,as a key component of visible light communication system,the nonlinear electro-optical transformation of LED and the memory effect brought by limited bandwidth restrict the signal transmission quality.In order to improve the performance of VLC system,it is necessary to suppress the nonlinear distortion in the transmission process.In view of the linear and nonlinear impairments,this paper models the VLC system and proposes two post distortion compensators based on machine learning.The main contents of this paper are as follows:Firstly,the visible light communication system is modeled.Using Wiener model,Rapps double-end saturation model and Lambert radiation model of point light source,we derive the channel response under direct link condition.The channel response includes LED nonlinear electro-optic conversion and memory effect.Then we model The VLC system by combining DCO-OFDM modulation.At the same time,the model is used to observe the changes of amplitude and constellation when the signals fall into different regions,so as to verify the reality of nonlinear distortion and the necessity of studying the nonlinear distortion compensation technique.Secondly,aiming at the disadvantage of the traditional post distortion method and decision tree method,a post distortion method based on XGBoost and grid search method is proposed to improve the fitting accuracy and generalization ability of the compensator.Simulation proves that under the condition of sacrifice certain computational complexity,our post distortion method based on XGBoost outperforms the methods based on linear recursive least square,Volterra series and memory polynomials.Compared to other compensator,the post distortion compensator based on XGBoost reduces the error vector magnitude and signal-to-noise ratio of the signal at high bias voltage.Besides,under the same bit error rate,it also increases the bias voltage and dynamic range of the signal.Finally,considering the before-and-after dependence of the signal caused by VLC system memory effect,we propose a post distortion method based on long-term and short-term memory(LSTM)network.By modifying the model input and using Adam optimization algorithm,the accuracy and convergence speed of the model can be improved.Compared with the post distortion compensator based on linear least square method,memory polynomial method and XGBoost method,the simulation results show that the post distortion compensator based on LSTM improves the transmission performance of the system under strong nonlinear electrooptical conversion and different baseband modulation.At the same time,based on the designed experimental system,the proposed LSTM post distortion compensator is used to improve the maximum transmission distance of the system under the condition of small driving voltage.