Low-Complexity Millimeter Wave CSI Estimation in MIMO-OFDM Hybrid Beamforming Systems

Channel state information (CSI) estimation in hybrid analog-digital (HAD) millimeter-wave (mmWave) massive MIMO systems is a challenging problem due to the high channel dimension and reduced number of radio-frequency chains. The problem becomes even harder when we consider wideband channels with higher frequency selectivity than the narrowband channels. Fortunately, by exploiting the sparse scattering nature of the mmWave channels and by adopting a simple setup at the transmitter, it was shown that the received signal can be organized into a third-order tensor that admits a Canonical Polyadic decomposition. Therefore, the channel parameters can be simply recovered once the decomposed factor matrices are estimated, e.g., using an alternating least square (ALS) method. However, ALS has a high computational complexity and a slow convergence rate. To resolve this issue, we propose a low-complexity noniterative algorithm to recover the channel parameters without requiring the estimation of the decomposed factor matrices by utilizing a compressed sensing technique and tensor algebra. Simulation results are provided to evaluate the effectiveness of the proposed algorithm.