Adaptive Reconstruction for Spatial Covariance Matrix in Hybrid Massive MIMO Systems

Spatial covariance matrix (SCM) plays an important role in super-resolution direction-of-arrival (DOA) estimation, which can be derived using traditional sample average approach. Hybrid analog-digital structure has been applied in massive multiple-input multiple-output (MIMO) systems for reducing the cost in millimeter-wave communications. However, the SCM cannot be obtained because the received signals at the antennas are unavailable to the digital receiver in hybrid massive MIMO systems. Based on our previous work, we will present an adaptive reconstruction approach in this paper where the SCM can be derived iteratively using the stochastic gradient descent algorithm. In this way, direct matrix inverse can be avoided so that the computational complexity for SCM reconstruction can be reduced significantly. Converge analyzes and numerical simulations are carried out to demonstrate the advantageous performance of the adaptive reconstruction approach.