Compressive Sensing Based Pilot Design for Spatial Correlated Massive Antenna Arrays

In this paper, we look at the raising spatial antenna correlations in massive antenna arrays and leverage spatial correlation combined with Compressive Sensing (CS) theory in the process of channel estimation. According to CS, the success probability of recovery is highly dependent on the restricted isometry property (RIP) of dictionary matrix. Recent advances in CS suggest that minimizing the coherence of dictionary matrix is an alternative efficient and effective way to test RIP. In this basis, this paper addresses the pilot pattern design problem in spatial domain aiming at minimizing the averaged coherence of the dictionary matrix. We first formulate an optimization problem with regard to pilot power distribution (PPD) and pilot antenna indexes set (PAIS) in CS-based channel estimation. Then two algorithms are proposed to separately design PPD and PAIS. Moreover, a jointly optimizing algorithm is presented. Simulation results demonstrate that the designed CS-based spatial pilot pattern outperforms random pilots and equal pilots, which significantly reduce pilot overhead and improve channel estimation quality compared with linear square (LS) estimation in spatial domain for massive antenna arrays.