Research On Compressive Sensing Channel Estimation In Massive MIMO System

The Massive MIMO system increases the degree of freedom by means of installing a large-scale antennas array at base stations and user equipment,and improves the performance deficiencies of the traditional MIMO system,such as insufficient capacity and low communication quality.Therefore,as the core technology of future communications,Massive MIMO theory once raised its research boom at home and abroad.To fully utilize the spatial multiplexing gains of Massive MIMO,accurate channel state information is required for signal coherent detection.However,as the number of antennas at both ends of the transmission and receiving terminal increases,the system also has difficulty in estimating the non-blind channel because the traditional method requires a lot of pilot overhead and resource waste in the Massive MIMO system.Therefore,it is hard to adopt the reconstruction algorithm which can reduce the pilot training overhead and improve the spectrum utilization when channel estimation is used in the Massive MIMO system.Compressive sensing technology applied to Massive MIMO systems can effectively reduce the pilot overhead required for channel estimation.The research shows that the channel matrixs tend to be sparse with the increase of the number of antennas in the Massive MIMO system,and the sparse channel can be reconstructed from a relatively small amount of measurement by using the compressed sensing techniques.This paper considers compressed sensing channel estimation in the mode of Frequency Division Duplexing(FDD)multi-user Massive MIMO system.Through analysis,the multi-user channel matrix is determined to have theBlock sparsity in the virtual angle domain and the joint sparsity feature including the common sparse part and the independent sparse part.Based on these two sparse features,a Joint Block Orthogonal Matching Pursuit(JBOMP)algorithm is proposed for channel estimation at the base station.The algorithm first sparsely represents the multi-user channel matrix in the virtual angle domain,then rewrites the channel transmission equation under the standard compressed sensing model by the sparsity of the channel,and finally rearranges the channel matrix according to the special sparse block structure feature of the Massive MIMO system,and establish the corresponding equivalent measurement matrix to reconstruct the channel.The performance of JBOMP algorithm is verified through simulation,which has better performance than the traditional algorithm.In order to further reduce the complexity of the JBOMP algorithm,this paper proposes two improved algorithms with low complexity.In the first algorithm,we only choice the users with high signal-to-noise ratio to find out the common sparse parts of all the users,and then iteratively find the independent sparse locations of all the users.In the second algorithm,different block lengths are used in estimating the common sparse locations and the sparse independent locations,thereby speeding the reconstruction time of the common sparse locations.The simulation results show that the two algorithms not only improve the system performance but also reduce the complexity in some scenarios.