Research On Intelligent Demodulation Technology Of Visible Light Communication Based On Machine Learning

Visible Light Communications(VLC)is a new type of wireless communication method in which light is used as an information transmission carrier and signals are loaded into visible light.It has larger communication bandwidth and higher communication speed,and has become a major part of communication field.At the same time,due to its advantages such as long transmission distance and good confidentiality,visible light communication also has very great application value in ocean and underwater communication.Signal demodulation plays a vital role in the VLC system.The existing methods for demodulation of visible light signals basically follow the coherent demodulation method in wireless communication,and most of the research on demodulation stays at the simulation stage assuming that the channel is pure Gaussian white noise channel,the traditional demodulator has good performance in this case.However,in the actual VLC link,affected by many uncertain factors such as channel environment,signal modulation format,background optical noise interference,etc.,the design of the demodulator in the optical receiver is often very complicated.Therefore,in view of the changes in the modulation format of the received signal and the channel environment,it is of important research value to realize a demodulator that does not need to know the priori information of the channel state in advance and has better demodulation performance.Machine learning has a good performance in fitting the channel characteristics of nonlinear systems.This paper introduces machine learning classification algorithm to the signal recognition at the receiving end of visible light communication,and research on the intelligent demodulation technology of visible light communication.The main work and research results of this paper are as follows:(1)An end-to-end visible light communication experimental platform was built according to the basic structure of the VLC system,and three common VLC modulation techniques,BPSK,QPSK,and 16QAM,were studied.Three formats of modulation signals were simulated on matlab,and the The visible light communication data transmission and sampling were successfully completed in indoor air and underwater channel environments,and the collected data was preprocessed to construct a VLC measured data set.(2)Modeling the demodulation system,transforming the signal demodulation problem in the VLC system into a classification problem in machine learning,and realizing three machine learning demodulators based on CNN,AdaBoost,and DBN-SVM.(3)Use the constructed VLC data set to train and verify the implemented machine learning demodulator model.Quantitative analysis of the performance of three demodulators in different modulation formats and different VLC channel environments The experimental results show that the machine learning demodulator has good demodulation performance in various modulation formats and channel environments.Among them,the AdaBoost demodulation algorithm has the best performance.For example,in an indoor air environment,for a 16QAM signal,when the signal-to-noise ratio is higher than 14dB,the demodulation symbol error rate can reach 10-3.DBN-SVM demodulation algorithm Secondly,the CNN demodulation algorithm has good performance at low signal-to-noise ratio,but the demodulation effect is not good at high signal-to-noise ratio.In addition,when comparing the demodulation performance in indoor air and underwater with alum added,it is found that the performance of the two links is close when the signal-to-noise ratio is low,but when the signal-to-noise ratio is higher than 12dB,The demodulation performance of the former is better than that of the latter by more than 4dB.