Research On The Technology Of Wireless Communication Interference Signal Identification And Processing Based On Deep Learning

Modern wireless communication system is facing increasingly complex electromagnetic environment,which is affected by many kinds of noise and interference signals.In the process of communication,if the communication partner can effectively identify the type of interference signal,it can take corresponding anti-interference measures to avoid or suppress the interference to the maximum extent and reduce the damage of interference to the communication quality.In recent years,with the development of deep learning and its outstanding performance in the fields of image and speech processing,it has been verified that deep learning has strong nonlinear mapping and data expression ability,which has stimulated researchers to apply deep learning to the field of communication anti-jamming and improved the anti-jamming ability of the system.In this dissertation,the technology of wireless communication interference signal identification and processing based on deep learning is studied as follows:The first chapter gives the background of this research,summarizes the current situation of interference identification and suppression in wireless communication,and gives the research content of this dissertation.The second chapter mainly introduces the structure of intelligent anti-jamming communication system and the mathematical model of typical repressive jam signal.The third chapter studies the interference identification network based on deep learning.In view of the problems that the interference identification algorithm based on feature extraction,which depends on the characteristic of artificial extraction,has the high complexity,and extracts features may incomplete or redundant,the real-value interference identification network is proposed.At the same time,in order to reduce the loss of signal phase information in the training of real-value interference identification network,a complex-value interference identification network corresponding to the structure of real-value interference identification network is proposed.Firstly,this part describes the interference identification architecture based on deep learning,and introduces the principle of real-value and complex-value CNN and the basic structure of residual network,and on this basis builds four interference identification networks: realvalue CNN,complex-value CNN,real-value ResNet and complex-value ResNet.Next,according to different input data formats and interference identification network structures,the interference signal identification performance is simulated and analyzed,and compared with the interference identification algorithm based on feature extraction.Then,from the three aspects of network structure,parameter complexity and convergence speed,the interference identification network based on deep learning is analyzed and compared.Finally,the migration performance of different interference identification networks is simulated and analyzed.The results show that the interference identification algorithm based on deep learning is better than the interference identification algorithm based on feature extraction.The overall performance of ResNet interference identification network is better than that of CNN interference identification network,which has faster convergence speed,but because of its deeper layers,the parameter complexity is higher.The overall performance of the complex-value interference identification network is better than that of the real-value interference identification network,which requires less parameters,faster convergence and greater generalization ability.Interference identification network has certain migration capability when the test set is distributed differently from the training set data.The fourth chapter studies the interference suppression network based on deep learning.In order to solve the problems existing in the traditional interference suppression,such as the high accuracy of parameter estimation in the interference reconstruction algorithm,the sensitivity of the interference detection algorithm to the decision threshold value,and the great influence of some interference elimination measures on the useful signals,the interference suppression network based on the complex-value U-Net is proposed.This chapter first gives the processing process of time domain and frequency domain interference suppression algorithm based on deep learning.Then,based on the signal time-frequency domain characteristics,a time domain and frequency domain interference suppression network based on complex-value U-Net is constructed.Finally,combined with the characteristics of signal waveform after interference suppression,the simulation analysis of the system BER after interference suppression is carried out in the single interference and mixed interference scene,and compared with the traditional typical interference suppression algorithm.The results show that the interference suppression algorithm based on complex-value U-Net can adaptively suppress the interference and recover the target signal by using the network's ability to express the data features,and the same network structure can handle multiple interference types,which is suitable for the scene with multiple interference.Finally,the fifth chapter summarizes the full text,and gives a brief description of the follow-up work.