Research On Channel Estimation For Millimeter Wave Massive MIMO Systems

The integration of millimeter-wave(mm Wave)and massive multiple-input multiple-output(MIMO)could provide higher date rates and bigger capacity.By accurately acquiring channel state information(CSI),mm Wave massive MIMO system can use beam selection to nearly achieve the Shannon Limit.By employing lens antenna array,the spatial channel can be transformed into beamspace channel,which could significantly reduce the dimension of MIMO system and the number of required RF chains.Therefore,this thesis investigated low-complexity beamspace channel estimation for wideband mm Wave MIMO-orthogonal frequency division multiplexing(OFDM)with lens antenna array system,optimized block-matching and 3-D filtering(BM3D)for mm Wave massive MIMO channel estimation and low-complexity massive MIMO channel estimation for mm Wave systems via matrix completion,respectively.Focused on the issue that the BM3D-based channel estimation scheme could not accurately estimate the channel for mm Wave massive MIMO systems,optimized BM3D-based channel estimation scheme for mm Wave massive MIMO systems was proposed.The key point for this channel estimation scheme was exploiting that the 3-D lens antenna array-based MIMO channel matrix can be regarded as a 2-D natural image.Utilizing the sparsity feature of the channel and the clustering feature of the paths,the BM3 D algorithm was optimized in order to improve the performance.Simulation results are provided to show that the proposed optimized BM3 D scheme can achieve satisfactory accuracy in all considered signal noise ratio regions.Focused on the issue in which the problem of high implementation complexity in wideband mm Wave MIMO-OFDM systems,a successive support detection channel estimation algorithm based on Gauss-Seidel(GS-SSD)method was proposed.Inspired by the successive support detection(SSD)scheme decomposed the total channel estimation problem into several channel components and used least square algorithm to acquire the matrix inversion,Gauss-Seidel method was used to approximate the matrix inversion which has high complexity.Simulation results and complexity analysis are provided to show that the proposed GS-SSD algorithm can significantly reduce the complex multiplication compared with SSD scheme and achieve high performance.Due to the inherent low-rank of the mm Wave channel,a fast randomized singular value thresholding(FRSVT)based channel estimation algorithm was proposed.By regarding the channel estimation as matrix completion problem,the proposed algorithm was based on the alternating direction method of multipliers with fast convergence properties and exploited FRSVT in order to significantly reduce the corresponding complexity.The simulation results and complexity analysis show that the proposed algorithm can achieve satisfactory channel estimation performance while maintaining low complexity.