Research On Channel Estimation Method Of GFDM System Based On Sparse Bayesian Learning

Generalized Frequency Division Multiplexing(GFDM)is a waveform scheme with great potential,which has many advantages such as high spectral efficiency,low out of band radiation,low peak average power ratio and flexible signal time-frequency structure.Accurate channel state information is crucial to GFDM symbol detection,however,the inherent self-interference in GFDM reduces Signal-To-Interference Ratio(SIR),which challenges channel estimation of GFDM systems under time-varying channel conditions.Therefore,it is of great significance for the development of GFDM to study GFDM channel estimation technology and reduce the number of pilots and improve the quality of channel estimation by utilizing the sparse characteristics of wireless channels.This thesis proposes a channel estimation algorithm for GFDM system based on Sparse Bayesian Learning(SBL).The specific work is as follows:(1)For time-invariant multipath Rayleigh channel,a GFDM channel estimation algorithm based on SBL is proposed.The sparseness characteristics of the channel are described by using hyperparameters,and the maximum likelihood estimation of the impulse response function of the sparse channel is calculated by applying the Expectation Maximization(EM)algorithm.The simulation results show that the proposed channel estimation algorithm based on SBL can accurately recover the channel state information by using the parameterized prior probability distribution of sparse channels,and the performance is significantly better than the traditional channel estimation method and the classical compressed sensing reconstruction algorithm.(2)For time-varying multipath Rayleigh channel,the multiple response signal model of GFDM system is constructed.Under the framework of Multiple Sparse Bayesian Learning(MSBL),Kalman filter and smoothing technology are used to track the channel,an SBL time-varying channel estimation algorithm(K-MSBL)for GFDM system based on Kalman filter and smoothing is proposed.Considering that the detection of GFDM data symbols can further improve the accuracy of channel estimation,a joint iterative channel estimation and symbol detection algorithm(JK-MSBL)based on Kalman filter and smoothing is proposed by using the estimation value of GFDM data symbols and pilot symbol information in GFDM frame.Simulation experiment proves that the algorithmic estimation performance of K-MSBL proposed by this research is significantly improved than the traditional MSBL due to the consideration of the correlation between the channel.The JK-MSBL algorithm,which further utilizes the data symbol estimation value,can obtain the Bit Error Rate performance close to the known channel condition,and has the characteristics of fast convergence and insensitivity to Doppler frequency shift.