Channel Estimation Based on Gibbs Sampling with π-Periodic Phase Only Measurements

This paper proposes a dropout and cluster deletion-K-means Gibbs sampling (DCD-KMGS) algorithm which solves the channel estimation problem in a halved-phase only single input single output (HPO-SISO) system, and its base station (BS) has the HPO radio-frequency (RF) chain use the phase detector to extract π-periodic phase measurements from the RF signals. Compared with traditional antennas, the HPO antennas have a simpler structure, lower power consumption and cost, which can alleviate the pain points in current fifth generation (5G) communication. However, the existing channel estimators cannot be applied into the HPO-SISO systems directly because the loss of phase information. The DCD-KMGS algorithm solves the problem by reconstructing the channel based on the phase measurements under the Bayesian framework. The core of DCD-KMGS algorithm is Gibbs sampling (GS) algorithm which can iteratively reconstruct the channel based on the real channel prior distribution and phase measurements. To obtain the channel prior distribution, the K-means clustering is applied to transform the real channel samples into the discrete prior distribution of the real channel, which can simplify the calculation of the channel prior distribution and the initialization of GS iteration. Then the damp, dropout and cluster deletion (CD) mechanisms are applied to increase the robust and lower the computational complexity. In the end, simulation experiments verify that DCD-KMGS algorithm correctly reconstructs channel in the HPO-SISO system and iterative gains.