Research On Channel Equalization Technology Of Atmospheric Laser Communication Based On LSTM Neural Network

With the rapid development of scientific research and the growth of engineering application demand,the requirements of laser communication technology for communication rate and BER are also increasing,and the complex atmospheric channel environment has a great impact on the data transmission performance of laser communication system,the laser signal in the atmospheric channel transmission process will produce light intensity flicker,beam drift and other phenomena,resulting in increased data transmission BER.To overcome these effects,laser signals are frequently processed by coding and equalization techniques,which cannot effectively improve the performance of communication systems in strong turbulent channels and cannot meet the demand for high rate data transmission under low BER conditions.In this paper,we address the need for optimizing the transmission performance of turbulent channels in atmospheric laser communication systems,combine the advantages of long short-term memory(LSTM)networks for time-series signals to carry out research,design a channel equalization method based on LSTM neural networks,and verify and analyze the proposed scheme through experiments.The main research works are as follows.1)A theoretical analysis of atmospheric laser communication and channel equalization is carried out.The basic principles of on-off keying(OOK)modulation and pulse position modulation(PPM)are introduced and the advantages of PPM modulation technique are explained.2)The neural network-based channel equalization technique for atmospheric laser communication is studied.Firstly,the basic principle of neural network is analyzed,the equalization method applicable to PPM modulated signal and the training optimization method of neural network equalizer are proposed,then the equalizer structure based on LSTM neural network is designed,the appropriate parameters are selected by analyzing the combination of activation function inside the equalizer,training algorithm and optimization algorithm,and finally the simulation of LSTM equalizer is carried out to verify the The theoretical analysis.3)Experiments on atmospheric laser communication equalization technology were carried out.Firstly,an atmospheric laser communication transmission system was built,and the atmospheric turbulence simulation device was used to simulate atmospheric turbulence of different intensities and analyze the effect of atmospheric turbulence of different intensities on the stability of signal optical power;then the data transmitted by atmospheric turbulence were collected for offline processing and a training method with random selection of samples was proposed to train the LSTM equalizer;finally,the BER performance of the LSTM neural network equalizer BER performance is verified and compared with LMS,CMA and DNN algorithms.The results show that using LSTM and DNN as equalizers can effectively reduce the BER,while the LMS and CMA equalization algorithms have gradually become insufficient to compensate for the effects of turbulent channels on optical signal transmission under medium to strong turbulence intensity.Using the LSTM neural network equalizer proposed in this paper reduces the BER by one order of magnitude under all intensities of turbulence than without equalization,and it is still possible to sacrifice a small amount of power to maintain low BER performance under moderately strong turbulence.