Signal Phase Noise Estimation And Optimal Decision Based On Deep Learning

In recent years,a new upsurge can be seen in the application of wireless networks which has greatly changed people's daily life.In the era of big data,people's demand for high-speed,high-capacity and reliable communication is increasing,but the spectrum resources are limited.In the increasingly congested spectrum environment,it is particularly important to achieve high-speed,and reliable communication in the limited bandwidth.High-frequency carrier,such as millimeter-wave and light wave,combined with high-order quadrature amplitude modulation(M-QAM),has been widely used as an effective method to greatly improve the communication rate.However,with the further increase of the transmission rate,the time slots and constellation become denser.The time-varying phase difference between the transmitter and the receiver has become the main obstacle to improve the communication rate.This kind of time-varying phase jitter is called phase noise.This paper mainly addresses the problem of reliable communication in the presence of uncertain phase noise.This paper first introduces the source of phase noise,establishes the information transmission model under the influence of phase noise,and explores the interference mechanism of phase noise on communication signals.Then we focus on the existing phase noise interference suppression methods,which can be divided into two categories.The phase noise suppression algorithm based on the transmitter is to shape the constellation to avoid rotation and overlap of constellation points due to the phase noise interference;at the receiver and demodulator,there is an approximate maximum likelihood detector based on the statistical characteristics of phase noise to reduce the symbol error rate.Then we make theoretical analysis and performance simulation of existing algorithms,analyzes and summarizes their merits and defects,and puts forward corresponding improvement schemes.Given the lack of theoretical support and poor efficiency of energy utilization of the existing constellation optimization methods,a joint optimization method of geometric shaping and probability shaping of constellation based on maximizing mutual information is proposed at the transmitter of communication system to improve the energy efficiency of constellation and avoid rotation and overlap of constellation points affected by phase noise.The receiver of communication system uses the joint optimization algorithm of channel parameter estimation and decision demodulation based on deep learning.In the joint optimization model,neural network is introduced to fit the posterior probability density of phase noise to solve the problem that it is difficult to calculate.At the same time,the estimation algorithm also provides the required noise parameters for maximum likelihood decision and solves the problem that the maximum likelihood decision algorithm requires noise parameters to ensure the information transmission rate and reliability under strong phase noise scenario.At last,this paper will start with the transmitter and receiver of the communication system jointly and carry out the co-simulation of the schemes proposed in this paper.The simulation results show that this co-optimization can improve the anti-interference ability of the communication system to the greatest extent.