Research On Channel Estimation Algorithm For OFDM System In The Time-frequency Double Selective Channel

Orthogonal Frequency Division Multiplexing(OFDM) technology is resistant to multipath interference and has a high spectrum utilization,which is used by many communication standards.Accurate channel estimation techniques are the basis of equalization and demodulation of OFDM receivers.With the advent of 4G and the rapid development of vehicles,we not only require superior communication experience in the slow time-varying scenario,but also hope to obtain good communication quality in a high-speed mobile scenario,however,fast movement of the terminal will cause the Doppler expansion,which will cause time-selective fading,as well as the multipath effect in the channel will cause frequency-selective fading.The traditional channel estimation algorithm can estimate the information of linear time varying channels,but it cannot solve the problem of too many parameters to be estimated in time-frequency double-selective channels,and the basis expansion model(BEM) uses the linear superposition combination of the basis function and its coefficient to fit a time-varying channel response,which can greatly simplify the number of parameters to be estimated.The main research contents and work of this paper are as follows:(1)The fading characteristics and classical simulation models of wireless channels are introduced.The implementation principle of each module in the OFDM system is briefly described,and base on this built a system-level simulation.(2)For the slow time-varing channel,principles of five traditional channel estimation algorithms are analyzed and their estimation performance of different channel models is compared by simulation.It is verified that performance of the DFT estimation is better than the LS estimation,and the complexity is lower than the MMSE estimation.Aiming at the problem that the traditional DFT estimation can only filters out the noise outside the CP and does not deal with the internal noise inside the CP,an improved algorithm is proposed in this paper,which set an appropriate threshold based on full consideration of the internal and external noise of the CP to maximize filter out the noise inside it.The simulation results show that the proposed algorithm has lower bit error rate than the traditional DFT estimation and its complexity does not increase.In addition,the commonly used interpolation algorithm and equalization algorithm are studied and compared by simulation.(3)For the fast time-varying channel,this paper models it by BEM,the principles of CE-BEM,GCE-BEM,P-BEM,DKL-BEM and DPS-BEM based on different basis functions are studied.The effect of Signal-to-Noise and Doppler shift on the estimate performance of each BEM by simulation.Aiming at the problem that the maximum frequency value of the basis function of CE-BEM is larger than the normalized maximum Doppler shift,an improved algorithm based on correcting the basis function coefficients is proposed in this paper,which can overcome the shortcomings that some frequency of basis functions are too high and the model estimation error is reduced.The simulation verifies that the mean square error performance of the proposed algorithm is lower than CE-BEM,which improves the channel fitting performance to some extent,but it is sensitive to Doppler shift.