Block Sparse Recovery for Wideband Channel Estimation in Hybrid mmWave MIMO Systems

Channel state information (CSI) is essential to achieve the optimal configuration of hybrid precoders and combiners in millimeter wave (mmWave) communication system. Exploiting the sparsity of mmWave channel enables to improve the CSI quality with small training overhead. In this paper, to further reduce the training overhead for channel estimation, the joint sparsity of wideband mmWave channel in angular-delay domain is exploited, where the mmWave channel estimation is formulated as a block sparse recovery problem. Accordingly, the block coherence of equivalent sensing matrix is smaller than the coherence of original sensing matrix, and decreases with the length of channel delay taps in the lower bound. The lower block coherence in turn elevates the recovery probability of unknown sparse channel. Finally, we proposed to employ the block orthogonal matching pursuit to exploit the derived block sparsity of wideband mmWave channel. The simulation results verify the analysis and demonstrate that the proposed block sparse recovery scheme outperforms the existing wideband mmWave channel estimators in terms of both the estimation accuracy and required training overhead.