Research On Demodulation Technology Of OAM Atmospheric Laser Communication System Based On Deep Learning

As the spectrum resources of microwave communication become more and more tense,people's requirements for communication speed and communication capacity are getting higher and higher.Free Space Optical Communications(FSO)has the advantages of large capacity,wide frequency spectrum,and strong anti-interference ability and gradually becomes the focus of people's research.In addition,the proposal of the Orbital Angular Momentum(OAM)beam brings a new modulation dimension to space optical communication,making the OAM-SK-FSO system attract the attention and research of a large number of scholars.However,for the OAM-SK-FSO system,the phase distortion and light intensity flicker caused by atmospheric turbulence,as well as the decrease in received power caused by pointing error and receiving aperture mismatch,make the traditional OAM demodulation method more difficult to demodulate.The accuracy rate of pattern recognition is reduced,and the communication error rate is increased.With the rise of machine learning technology,Convolutional Neural Networks(CNN)algorithms can mine various features of objects at a deep level.For the OAM spot pattern,the convolutional neural network can correctly identify the OAM mode according to the spiral phase characteristics of different OAM modes,which provides a new idea for the demodulation of the OAM signal.This paper studies the demodulation performance of OAM mode of OAM-FSO communication system by convolutional neural network under the influence of atmospheric turbulence,pointing error and receiving aperture mismatch.The main research work and results are as follows:Firstly,starting from the theory of Kolmogorov,this article makes a detailed study of atmospheric turbulence,useing different turbulence models to quantify atmospheric turbulence at different intensities,and useing power spectrum inversion and Zernike polynomial methods to conduct atmospheric turbulence.In addition,the influence of pointing error on the received beam is studied by modeling the cross-section of the received beam.Secondly,this paper design and built an OAM-SK-FSO communication system.The propagating beam is OAM-SK modulated by the spatial light modulator,and five different OAM modes are modulated onto the propagating beam,representing five different data symbols,and a phase screen that simulates atmospheric turbulence is loaded during the propagation process.Finally,at the receiving end,the CCD camera is used to sample and receive,and the received light spot is demodulated using the CNN algorithm.Third,under the influence of atmospheric turbulence,pointing error,and receiving aperture mismatch,the demodulation accuracy of the 8-layer CNN based on the convolutional neural network algorithm for the OAM mode is studied.The experimental results show that when the pointing error is zero and there is no receiving aperture mismatch,the CNN can recognize the OAM mode with 100%accuracy under weak turbulence,and the recognition accuracy can reach 91.5%under strong turbulence.When there is a pointing error,the CNN can still achieve 100%correct recognition of OAM under weak turbulence.When the receiving aperture is mismatched,the recognition accuracy of OAM can reach 100%under weak turbulence.When the pointing error and the receiving aperture mismatch occur at the same time,CNN can also identify the OAM mode with 93%accuracy under weak turbulence.