Low-rank channel estimation for mm-wave multiple antenna systems using joint spatio-temporal covariance Matrix

Millimeter-Wave (mm-Wave) and very large multiple antenna systems (VLMAS) are two key technologies in the deployment of Fifth Generation (5G) mobile communication systems. In order to exploit all the benefits of VLMAS, spatial

and temporal (ST) features must be estimated and exploited to compute the precoding/decoding matrices. In the literature, a practical channel estimation approach is proposed by assuming that the spatial features are completely unknown, leading to nonparametric estimation in which antenna array calibration is not required. Additionally, when the signal-to-noise ratio (SNR) is not so high, a low-rank (LR) version of the estimated channel is proposed that provides better performance than the full-rank (FR) one in terms of bias-variance trade-off in the mean squared error (MSE). However, previous work assumes that spatial and temporal characteristics of the channel can be estimated separately. Then, the performance is degraded in realistic channels. In this paper, we propose an alternative way to characterize the FR estimated channel using a joint ST covariance matrix, combined with a low-complexity semi-parametric spatial response and delay estimation technique. Moreover, we propose an automatic rankselector (ARS) based on the MSE in order to provide the best LR channel estimation for each scenario. Numerical results show that the proposed technique outperforms existing approaches in the literature.