Downlink Channel Estimation And Pilot Design Based Channel Sparse Representation For Massive MIMO Systems

With the rapid increase in the number of smart terminals and the demand for mobile services,existing wireless communication technologies are facing new challenges.Large scale multiple-input multiple-output(MIMO)technology is one of the core technologies to achieve breakthrough in the field of mobile communication,which greatly improves the power efficiency and spectrum efficiency of wireless communication system by equipping large-scale antenna array on the base station side.In order to obtain the potential capability brought by massive MIMO,accurate channel state information(CSI)acquisition is essential.In the conventional channel estimation method based on orthogonal pilot assistance,the pilot cost is affected by the size of the antenna array on the base station side,and the larger the number of antennas on the base station side,the larger the pilot cost.This paper focuses on the pilot extension problem and channel estimation performance problem in the massive MIMO system.It studies the downlink transmission pilot design and channel estimation problem,and the specific contents are as follows:First,this paper reviewed several conventional pilot-assisted channel estimation methods in the downlink of massive MIMO systems,including least squares(LS),and linear minimum mean square error(LMMSE)and Relaxed Minimum Mean Square Error(RMMSE)estimation algorithms.The relationship between the channel estimation algorithm and the number of antennas on the base station side is analyzed by mean square error.Finally,the theoretical results are verified by simulation.Secondly,in order to solve the problem of excessive pilot overhead of conventional channel estimation algorithms in massive MIMO systems,a channel estimation method based on channel sparse representation is proposed.Massive MIMO channels can be sparsely decomposed in the beam direction.In the original uniform linear array(ULA),the beam domain channel can be obtained by discrete fourier transform(DFT)matrix sampling.There was a problem with channel energy leakage in this method.For this reason,it is proposed to use two kinds of redundant basis to finely sample the channel,the channel coefficients obtained are more sparse and the energy is more concentrated.Based on this,a beam domain pilot matrix that satisfies the constraint equidistance is designed,and it is verified that it also meets the unique sparse recovery condition under the redundant basis.Finally,the sparse bayesian learning(SBL)algorithm is improved based on the relationship between the redundant basis components and the noise variance to solve the channel estimation problem under unknown statistical channel information.Simulation results show that the performance of this channel estimation scheme is significantly better than that of LS estimation under orthogonal pilots and close to MMSE estimation.Finally,a channel estimation method based on channel sparse representation is proposed to solve the problem of excessive pilot overhead in a massive MIMO orthogonal frequency division multiplexing(OFDM)system.Massive MIMO-OFDM channels can be decomposed in the beam and delay directions,and energy is concentrated in the first few taps in the delay direction.By combining the sparse channel sampling in the beam direction and the delay direction,a beam delay domain channel with higher sparse characteristics and better energy concentration can be obtained.In view of the energy leakage problem of DFT matrix sampling,the overcomplete DFT matrix is used instead of the DFT matrix to finely sample the channels,thereby improving the sparseness of the channel coefficients.Then,using the properties of zadoff Chu(ZC)sequence,a deterministic pilot scheme based on multi root phase shifted ZC sequence is proposed,and on the basis of multiple measurement vector-SBL(MSBL)algorithm,a channel estimation method based on multiple received signal samples and MSBL(MSCE)algorithm is proposed.The simulation results show that the performance of the channel estimation scheme is better than that of LS estimation,and it is close to MMSE estimation at high SNR.