Research On Compressed Sensing Based Millimeter Wave Channel Estimation Algorithms

Owing to its large bandwidth,millimeter wave(mmWave)communication will be one of the important technologies of the next generation wireless communication systems.Since many essential technologies,such as precoding/combing,are dependent on channel state information(CSI),it is crucial to realize mm Wave channel estimation with high accuracy and low training overhead.For one thing,some hardware constraints,such as limited number of radio frequency(RF)chains and limited resolution of phase shifters,make it difficult to estimate the channel precisely.For another,the numerous antennas equipped in mmWave systems obstruct the task of reducing the estimation overhead.Confronting these challenges,this thesis studies the compressed sensing based mmWave channel estimation algorithm,expecting a more precise and efficient channel estimation.First of all,as for hybrid beamforming architecture with phase shifters,this thesis studies power based adaptive compressed sensing scheme,which relies on codebook design to a large extent.Traditional codebook design using least squares fails to generate beams that cover expected spatial angular intervals.So we propose iterative adjustment(IA)to improve codebook design,followed by the proof of its convergence.We show that in theory,enhanced codebook design can force beams divide the spatial angular intervals precisely as the requirements of IA.In addition,it is necessary to decompose the ideal codes into analog and digital ones in hybrid beamforming systems.When the number of RF chains is relatively low,the decomposition by orthogonal matching pursuit(OMP)will cause severe distortion of beams.To resolve this problem,this thesis devises the codebook design based on generic algorithm,which is capable of obtain more expected beams.Simulation results show that the two improvements on codebook reduce the estimation error of spatial angular effectively and thus improve the spectral efficiency of mm Wave systems.Second,this thesis proposes a novel algorithm,correlation based adaptive compressed sensing algorithm.The proposed algorithm can obtain precise spatial angular estimation by correlation based detection in each level,which is not constricted by the resolution of beams as in the power based method.For this algorithm,this thesis also proposes two criteria for beam pattern design and gives one suitable beam pattern,cosine beam pattern.Considering practical hybrid beamforming architecture and limited-resolution phase shifters,this thesis designs hybrid codebook of this cosine beam pattern.Two steps are needed to generate the hybrid codebook:the first step is generating basic codes of each level using symmetrical-OMP algorithm and the second step is acquiring all the codes of this level by code rotation.Simulation results show that the proposed schemes is able to achieve less wrong estimation probability even with lower training overhead,and thus achieve higher spectral efficiency.Finally,this article studies channel estimation algorithm in mm Wave systems with 1-bit analog to digital converter(ADC).In this kind of systems,large scale antenna array is employed in base station(BS)and each moving station uses single-antenna and receives signal with 1-bit ADC.In terms of this system,this thesis gives the compressed sensing model of estimation problem.Then,BIHT algorithm and GAMP algorithm are used to solve this problem respectively.This thesis gives a complete and detailed interpretation of the derivation of GAMP,and then derives the output function in the case of 1-bit sampling of additive Gaussian output.The convergence curves of the two algorithms show that they both can converge quickly.The simulation results of the mean squared estimation error(MSE)corresponding to signal to noise ratio(SNR),path number and training overhead show that comparing with BIHT,GAMP has smaller MSE,requires lower training overhead and is rarely affected by path number.