A Two-Stage Estimator for Time-Varying Millimetre Wave Channel via Covariance Matrix

In high-speed mobile scenarios, conventional estimation schemes for millimetre-wave multiple-input multiple-output (MIMO) communication systems may fail to track of time varying channels, so this paper proposes an efficient channel estimation scheme for time-varying millimetre-wave channels. Since the arrival/departure angles (AoAs/AoDs) vary much slower than the path gains, the channel estimation is decoupled into two stages. In the first stage, we utilise a full-rank pilot structure to construct a maximum likelihood estimation (ML) of the observed diagonal covariance matrices to reduce the inter-user interference and estimate the channel covariance matrix. Then, we utilize the special structure of the channel covariance matrix to accomplish the estimation of AoAs/AoDs, Doppler shift with low complexity. In the second stage, we utilize the estimates from the previous stage as a priori knowledge and track the time-varying parameter path gains using a Kalman filter to achieve accurate estimation of the path gain. Simulation results show that the proposed scheme can accurately estimate the time-varying millimeter-wave channel based on a small number of pilot frequencies.