| Massive Multiple Input Multiple output(MIMO)technology can bring higher spectral efficiency and power efficiency,thus being one of the directions in the fifth generation with highest potential.Channel estimation is important by improving the system performance,user scheduling and transmission scheme.Besides,Channel estimation and scheduling,commonly contain high-dimension matrix computation,thus,Channel estimation and transmission scheme optimization in massive MIMO wireless system have a high complexity.By configuring a large number of antennas(100-1000)in the base station the energy can be concentrated in a small space to improve the throughput,also the antenna makes a full use of the space using the same frequency resource at the same time and multiple terminals(16-64)concurrent communication to reduce overall energy efficient,nonetheless the premise of using this characteristics is that the base station must estimate the downlink channel specially when the number of antennas increase the traditional channel estimation scheme will bring huge pilot overhead,and channel estimation becomes almost impossible.This Thesis will mainly discuss the downlink non-blind channel estimation algorithm under massive MIMO,the MIMO model adopts multi-user MIMO and works in FDD mode.The innovation of this Thesis has the below mentioning points:(1)An improved algorithm for structured compressed sensing with adaptive sparsity is proposed for wireless channels this feature of sparsity can be used to reconstruct the channel matrix through compressed sensing technology.Subspace Pursuit(SP)algorithm in the traditional algorithm has a backtracking mechanism,which makes its signal reconstruction accuracy higher,However like other algorithms,the SP algorithm requires signal sparsity as a prior condition,and It is difficult to obtain sparsity in some environments,so this algorithm is not suitable for massive MIMO environment.The Adaptive Sparsity Matching Pursuit(ASMP)algorithm which based on the improvement of the SP algorithm adds a sparsity adaptive mechanism,so ASMP is more practical,Although ASMP can perform channel estimation adaptively,as the number of antennas at the BS continues to increase the algorithm pilot overhead will increase,and its performance cannot meet actual needs.Combining these characteristics with the ASMP algorithm and adopting a non-orthogonal pilot structure,Then the Block based Adaptive Sparse Matching Pursuit(BASMP)algorithm is proposed.The experiments show that the proposed BASMP algorithm can not only adapt to sparsity as ASMP,but also more suitable for massive MIMO environment than ASMP algorithm.(2)The Information Set Decoding(ISD)algorithm is an existing improved convex optimization algorithm.Therefore,this paper proposes an improved channel estimation algorithm based on the ISD algorithm.Channel estimation is achieved by transmitting non-orthogonal pilots.The proposed Structured Iterative Support Detection(SISD)algorithm first solves the basis pursuit(BP)problem by alternating squaring and using the alternating direction method(ADM),updates the initial signal support set by a threshold detection strategy,and finally performs alternately through three processes,the signal structure will determine the final support set and get a sparse channel matrix estimate.The experiments show that the improved algorithm can achieve ideal channel estimation results by updating the support set iteratively.(3)An improved channel feedback algorithm is proposed to reduce feedback overhead.The user acquires downlink CSI after that it needs to be fed back to BS as a signal through the uplink channel for beamforming,resource allocation and other processes.Nowadays,feedback schemes based on compressed sensing technology are more commonly used,but as the number of antennas in Massive MIMO increases this will lead to excessive feedback overhead,and the feedback link bandwidth is limited,that’s why a downlink channel based on differential operation estimation and feedback algorithm is proposed,which first use compressed sensing technology to estimate the downlink channel matrix at the user terminal,and then Performs differential operations on any two adjacent symbols and the differential value of channel matrix will be obtained.Finally through the uplink the difference value will fed back to the base station as signal compression.The experiment shows that it is feasible. |
