Hierarchical Multi-Beam Search Based ChannelEstimation for Millimeter-Wave Massive MIMO Systems

Accurate channel state information (CSI) acquisition is indispensable for mmWave massive MIMO systems to perform sophisticated signal processing. The existing channel estimation methods such as beam training and compressed sensing either have low accuracy or require considerably high training overhead. To achieve a fast yet accurate channel estimation, we propose a parallel hierarchical multi-beam search method in this paper. Unlike the existing beam search methods, which search multipath components (MPCs) one by one, the proposed method searches all the MPCs simultaneously by scanning all the beam pairs with MPCs in each layer of codebook in parallel. Therefore, the beam search of different MPCs does not interfere with each other. To improve the estimation accuracy and reduce the beam training overhead, the length of each training sequence under different codebook layers and different SNR is obtained by beam detection. Moreover, we particularly design an optimal beam pattern synthesis codebook for the hierarchical multi-beam search by using the design method of digital filter, which is applicable in other hierarchical search methods as well. Simulation results show that the proposed codebook is superior to other codebooks and the proposed multi-beam search method has higher channel estimation accuracy, but requires less training overhead, compared with the state-of-the-art alternatives.