An Improved Beam Training Scheme Under Hierarchical Codebook

In millimeter wave (mmWave) communications, the beamforming gains of large-scale antenna arrays are used to combat severe path losses and improve communication link quality. This requires beam training (BT) to determine the optimal beam direction and achieve beam alignment. The hierarchical codebook with multi-resolution training beams is widely used in BT to reduce training overhead. Unfortunately, due to hardware restrictions, the transition area between two adjacent training beams often leads to error propagation in a traditional BT scheme, which limits the accuracy of beam alignment. In this paper, we propose an improved BT scheme under hierarchical codebook. First, we define an index to measure the power difference of the received signals. Once the index is lower than the predefined threshold, the optimal beam direction is considered to be in the transition area, and an extra interval that contains the transition area is proposed for further beam searching. Then, we analyze the optimal threshold and the probability of beam misalignment based on the ideal beam pattern, proving that the error rate exponentially decreases with the signal-to-noise ratio (SNR) under a single-path channel with constant gain. Finally, simulation results demonstrate that the proposed scheme can significantly improve the accuracy of beam alignment and is applicable to most existing beam patterns.