Estimation of Block Sparse Channels with Conjugate Gradient SAMP in Massive MIMO Systems

The hyperscale channel matrix of massive multiple input multiple output (MIMO) systems increases the difficulty of channel estimation. However, the sparsity of channel matrix brought by the scale expansion opens up a new path for channel estimation. The sparsity adaptive match pursuit (SAMP) algorithm still has the reconstruction ability even when the sparsity of channel is unknown. But there are a large number of inversion operations in the calculation process, which results in computational inefficiency. The conjugate gradient (CG) method has better efficiency and precision for large sparse matrix operation. Therefore, the classical SAMP algorithm is improved in this paper and a block sparsity based conjugate gradient SAMP (BCG SAMP) algorithm is proposed. Vectorization transformation is utilized to extract the block sparsity of the channel and regularization and CG methods are further used to estimate the channel, thus higher efficiency of the algorithm can be achieved. Simulation results show that the proposed BCG SAMP algorithm can increase the reconstruction probability, precision and time efficiency under certain accuracy requirements.