Noise variance estimation for 5G wireless networks under pilot contamination

Large-scale multi-cell multi-user multiple-input multiple-output (MU-MIMO) is expected to be one of the enabling technologies for fifth generation (5G) time division duplexing (TDD) systems. One of the major challenges faced by this technology is the pilot contamination issue where interference from users in the neighboring cells may significantly impact the performance of the channel estimation process. There are different strategies proposed in the literature to overcome the pilot contamination issue. However, most of these works assume perfect knowledge of the noise variance in the channel estimation process which is not the case in realistic systems. In this paper, we study the pilot contamination issue in multi-cell MU-MIMO cellular networks and propose two noise variance estimators; 1) maximum likelihood, and 2) method of moments estimators which can be used for the channel estimation under the pilot contamination. We evaluate the performance of these noise estimators, and the impact of the noise estimation on the channel estimation performance. Further, the existing literature shows that the effects of the pilot contamination on the channel estimation is vanished when the angle-of-arrival of the desired and interfering users do not overlap. Using a pilot assignment strategy from the existing literature, we evaluate channel estimation performance at mmWave frequencies for large antenna array regime considering recently developed 5G channel models. Our simulation results show that under pilot contamination, better channel estimation can be achieved at mmWave bands compared to low frequency scenarios.