The problem of pilot contamination reduction in massive MIMO for physical channel models

In this paper, we consider massive multiple-input multiple-output (MIMO) multicell systems with very large antenna arrays at base station (BS). We propose an eigenvalue decomposition (EVD) approach to evaluate the covariance matrix of the received vector at the BS, and focus on a physical channel model where the angular domain is partitioned into a finite number of directions, so-called angle of arrival (AOA). In this scheme, the asymptotic orthogonality of fast fading coefficients between users and the BSs would be true only when the number of BS antennas goes to infinity. Furthermore, we derive an optimal multiuser detector by taking the channel estimation and the intercell interference from other cells into account to decrease channel estimation error, hence, the system performance will be enhanced significantly, even in strong inter-cell interference environment. Numerical results are provided to verify our analysis.