Sparse/dense channel estimation with non-zero tap detection for 60-GHz beam training

Estimation of the multipath channel in 60-GHz communications is challenging, because the channel may be sparse or dense during beam training. Specifically, because of the variation of the number of non-zero taps, it is hard for common estimators to obtain robust and prominent performance. In order to address this problem, the authors propose a sparse/dense channel estimation with non-zero tap detection (SDCE-NTD). The estimation is conducted in a three-stage fashion, including initial estimation with the unstructured least-square (LS) algorithm, non-zero-tap detection with the generalised likelihood ratio test approach, and posterior estimation with the structured LS algorithm. The false-alarm and detection probability of the tap detector, as well as the mean square error (MSE) of SDCE-NTD, are derived and confirmed via simulations. Comparisons are conducted between SDCE-NTD and the common estimators in the beam training scenarios, where both dense and sparse channels exist. Results show that SDCE-NTD reveals a significant gain in terms of MSE over both the conventional LS algorithm, which does not exploit the sparse nature of the channel, and the matching pursuit algorithm, which endeavours to exploit the sparsity. In addition, it is also demonstrated that the proposed estimator can approach the lower bound with high signal-to-noise ratio.