| Driven by both market demand and technological progress,wireless communication systems experience a technological revolution approximately every ten years.The 5G system achieves higher system capacity than the previous generation 4G system.The performance of beamforming and spatial multiplexing in massive MIMO systems depends on the availability of channel state information.Effective acquisition of channel state information can achieve better capacity.Although extensive research has been done on different channel state information acquisition methods of traditional MIMO systems,in millimeter wave communication,due to the high frequency and short wavelength of millimeter waves,the part of the transmission process that is shielded by objects and the scattering part has a large loss,and the line-of-sight Transmission is the main propagation mode,and traditional methods cannot be directly applied to millimeter wave systems.Therefore,research on channel estimation algorithms in millimeter-wave Massive MIMO is of great significance.The main research content of this thesis is as follows:The wireless channel in the millimeter-wave massive MIMO system has sparse characteristics,and the convex optimization algorithm is an effective means to solve the channel estimation problem.By introducing the Nesterov smoothing process,the objective optimization problem is transformed into a form that is light to get gradients.On the ground of this objective optimization issue,a quick iterative accelerated iterative algorithm is proposed.One of the key ideas of the algorithm is the weighted average of the iterative sequence,this process can improve the convergence characteristics of the standard gradient descent algorithm,and through the fixed point continuous process,the target problem is turned into a series of small problems,and each time the optimization work is done,the threshold is cut to speed up the algorithm convergence.The simulation results point out that the designed algorithm has a great performance improvement compared with the orthogonal matching algorithm under the condition of high signal-to-noise ratio,and has faster convergence speed,less iterations and running time than other iterative algorithms.Aiming at the redundant iterative problem caused by artificially setting parameters in the iterative algorithm used in the sparse signal recovery problem in the channel estimation of the mm Wave Massive MIMO system,the sparse linear inverse problem is solved by expanding the iterative algorithm into a neural network.The number of iterations is reduced by setting parameters such as threshold as learnable parameters,and the influence of the selection of iterative algorithm and the setting of learnable parameters on the performance of iterative neural network is analyzed.The simulation results show that compared with the original iterative channel estimation process,the neural network with learnable parameters using the same iterative algorithm can obtain better estimation performance with fewer iterations.And the neural network designed by the faster convergence algorithm has better performance. |
