(A) study on statistical sparse recovery for detection and channel estimation in communication systems

Abstract

In this dissertation, we delve into Bayesian sparse recovery to enhance sparse recovery performance and reduce computational complexity in wireless communication systems. To address the significant computational complexity of existing sparse recovery algorithms, we explore a random likelihood decoding technique based on Markov chain Monte Carlo sampling. We utilize Metropolis-Hastings sampling for support identification and Gibbs sampling for sparse signal estimation. Specifically, we propose a two-stage algorithm that iteratively conducts both sampling processes and the entire process to improve the performance of sparse vector recovery. Additionally, with next-generation 6G communication systems incorporating more antennas and higher frequency bands, there is an anticipation of greater utilization of the near-field region compared to traditional communications that predominantly focus on the far-field region. To address this shift, we propose a pilot signal and channel estimator co-design technique aiming to optimize sparse recovery performance in hybrid-field extremely large-scale MIMO systems where both far-field and near-field regions coexist. Specifically, we propose an alternating direction method of multipliers (ADMM)-based pilot signal design algorithm that can maximize sparse vector recovery performance and a hybrid-field channel estimation algorithm using Bayesian approaches based on the designed pilot signal.