Research On The Channel Estimation And Equalization Technology Based On Deep Learning

Channel estimation and equalization technology is an important part of the communication systems,and its performance directly affects the accuracy of the received information.In the traditional channel estimation algorithm based on time-domain noise reduction,the channel estimation performance of the system is poor because the noise location information cannot be accurately obtained.Meanwhile,the accuracy of channel frequency information will also affect the transmission reliability of the system when the iterative equalization algorithm is used.In recent years,there has been significant achievements for deep learning(DL)in computer vision and natural language processing.Also,DL has a huge potential in the wireless communicatuion field,which has been widely concerned by academia and industry.Therefore,aiming at the problem of channel estimation and equalization in wireless communication systems,based on the traditional algorithms,this thesis studies the channel estimation and equalization algorithms based on deep learning,so as to provide reference for the design of future wireless communication systems.The main work of the thesis is:Firstly,the thesis introduces the related concepts of wireless channels and then analyzes the changes of channel frequency-domain state information in three typical scenarios: extended pedestrian a model(EPA),extended vehicular a model(EVA)and extended typical urban model(ETU).Finally,the thesis illuminates the basic principle of orthogonal frequency division multiplexing(OFDM)system,the related concepts of deep learning and the training methods of the neural network model.For the optimization problem for the least square estimation algorithm in OFDM,an improved DFT channel estimation algorithm based on deep learning is proposed by analyzing the traditional channel estimation algorithm based on discrete fourier transform(DFT)and the adaptive time-domain filtering algorithm based on neural network.This algorithm builds a fully connected neural network model to get accurate multipath location information and filter the noise of non-multipath location in the cyclic prefix,so as to improve the performance of channel estimation.Simulation results show that compared with the traditional channel estimation algorithms,the proposed algorithm has lower mean square error(MSE)and bit error rate(BER)and better channel estimation performance in the EPA,EVA and ETU scenarios.For the optimization problem for the frequency domain iterative equalization technology in OFDM,by analyzing the principle of the traditional zero forcing equalization(ZF),minimum mean square error(MMSE)and MMSE-turbo frequency domain iterative equalization algorithm,an iterative equalization algorithm based on deep learning is proposed.The algorithm can obtain more accurate channel frequency-domain information and better BER performance by building a convolutional neural network model to optimize the input frequency-domain state information.Simulation results verify the effectiveness of the proposed algorithm.