Channel Acquisition for Hybrid Analog-Digital mMIMO System by Exploiting the Clustered Sparsity

This paper considers the channel estimation for massive multiple-input multiple-output (mMIMO) systems, where the base station (BS) employs hybrid analog-digital processing to reduce the hardware complexity. With hybrid analog-digital processing, the BS cannot obtain enough training samples since the number of radio frequency (RF) chains is limited. This results in significant difficulty in the design of channel estimation scheme. This paper aims to address this problem using the emerging machine learning techniques. By exploiting the sparsity property of angular-domain channel, a novel clustered sparse Bayesian learning (SBL) approach is proposed to estimate the channel reliably. The scheme does not require the knowledge of channel statistics and angular information of users. The numerical simulations show that the proposed scheme based on clustered SBL outperform the reference schemes significantly in term of normalized mean square error (NMSE).