Multidimensional Sparse Recovery for MIMO Channel Parameter Estimation

TL;DR

This paper introduces a novel sparse recovery method for MIMO channel parameter estimation that decomposes a multidimensional problem into simpler one-dimensional problems, improving efficiency and accuracy.

Contribution

The paper proposes a decomposition-based sparse recovery approach using nuclear norm minimization and STELA algorithm for better multidimensional channel parameter estimation.

Findings

Effective decomposition into 1D problems

Reduced off-grid sensitivity

Efficient implementation with STELA

Abstract

Multipath propagation is a common phenomenon in wireless communication. Knowledge of propagation path parameters such as complex channel gain, propagation delay or angle-of-arrival provides valuable information on the user position and facilitates channel response estimation. A major challenge in channel parameter estimation lies in its multidimensional nature, which leads to large-scale estimation problems which are difficult to solve. Current approaches of sparse recovery for multidimensional parameter estimation aim at simultaneously estimating all channel parameters by solving one large-scale estimation problem. In contrast to that we propose a sparse recovery method which relies on decomposing the multidimensional problem into successive one-dimensional parameter estimation problems, which are much easier to solve and less sensitive to off-grid effects, while providing proper parameter pairing. Our proposed decomposition relies on convex optimization in terms of nuclear norm minimization and we present an efficient implementation in terms of the recently developed STELA algorithm.