Nonconvex Regularized Gradient Projection Sparse Reconstruction for Massive MIMO Channel Estimation

Novel sparse reconstruction algorithms are proposed for beamspace channel estimation in massive multiple-input multiple-output systems. The minimize a least-squares objective having nonconvex regularizer. This regularizer removes the penalties on few large-magnitude elements from conventional $\ell _{1}$ -norm regularizer, and thus it only forces remaining that expected to be zeros. Accurate fast reconstructions can achieved by performing gradient projection updates within framework of difference convex functions (DC) programming. A double-loop algorithm single-loop via different DC decompositions, these two have distinct computational complexities convergence rates. An extension is further designing new step sizes algorithm. has faster rate achieve approximately same level accuracy as Numerical results show significant advantages over existing terms accuracies runtimes. Compared with benchmark approaches, smaller error higher achievable spectral efficiency.