Time-frequency doubly selective channel estimation based on compressed sensing

In this paper,considering time-frequency doubly selective channel,we utilize the channel's time correlation that the channel coefficientscorresponding tothe neighboring instants have a strong correlation.And we present a linear approximation method,which effectively reduces the number of unknown parameters.Considering the sparseness of the wireless channel in the delay domain,this paper reconstructs unknown channel parameters of the proposed model based on compressed sensing (CS) theory.In the simulations,we observe the system performance of the linear approximation model and the non-linear approximation model,respectively,and present the normalized mean squared error (NMSE) curves based on the least square (LS),orthogonal matching pursuit (OMP) and sparse Bayesian learning (SBL) algorithms.Simulation results show that the linear approximation method can effectively model the time-frequency doubly-selective channels.For the proposed linear approximation model,SBL algorithm can accurately recover the channel response,and overcome the Doppler effecteffectively.