[Paper Review] Beamspace Channel Estimation in mmWave Systems via Cosparse Image Reconstruction Technique

Jie Yang, Chao-Kai Wen+2 more|arXiv (Cornell University)|Jul 28, 2017

Millimeter-Wave Propagation and Modeling34 references5 citations

TL;DR

This paper proposes a novel beamspace channel estimation technique for 3D lens antenna array mmWave systems using the SCAMPI algorithm enhanced with Gaussian mixture probability modeling via EM learning. By exploiting the sparsity and smoothness of the beamspace channel as a 2D image, the method achieves higher accuracy and faster convergence than OMP and support detection, with robustness to 10% phase shifter reduction and improved performance over uniform prior models.

ABSTRACT

This paper considers the beamspace channel estimation problem in 3D lens antenna array under a millimeter-wave communication system. We analyze the focusing capability of the 3D lens antenna array and the sparsity of the beamspace channel response matrix. Considering the analysis, we observe that the channel matrix can be treated as a 2D natural image; that is, the channel is sparse, and the changes between adjacent elements are subtle. Thus, for the channel estimation, we incorporate an image reconstruction technique called sparse non-informative parameter estimator-based cosparse analysis AMP for imaging (SCAMPI) algorithm. The SCAMPI algorithm is faster and more accurate than earlier algorithms such as orthogonal matching pursuit and support detection algorithms. To further improve the SCAMPI algorithm, we model the channel distribution as a generic Gaussian mixture (GM) probability and embed the expectation maximization learning algorithm into the SCAMPI algorithm to learn the parameters in the GM probability. We show that the GM probability outperforms the common uniform distribution used in image reconstruction. We also propose a phase-shifter-reduced selection network structure to decrease the power consumption of the system and prove that the SCAMPI algorithm is robust even if the number of phase shifters is reduced by 10%.

Motivation & Objective

• To address the high hardware complexity and training overhead in mmWave systems with large antenna arrays.
• To exploit the energy-focusing capability of 3D lens antenna arrays and the sparsity of beamspace channels for efficient estimation.
• To improve upon existing compressive sensing algorithms like OMP and support detection by modeling channel distribution with a Gaussian Mixture (GM) prior.
• To reduce system power consumption by proposing a phase-shifter-reduced selection network.
• To demonstrate robustness and performance gains under practical hardware constraints, such as reduced RF chains and phase shifters.

Proposed method

• The beamspace channel matrix is modeled as a 2D natural image with sparse, smoothly varying entries, enabling application of image reconstruction techniques.
• The SCAMPI algorithm—based on sparse non-informative parameter estimation—is adopted for fast and accurate channel recovery.
• A Gaussian Mixture (GM) probability model is introduced to represent the channel distribution, replacing the common uniform prior.
• The EM learning algorithm is embedded into SCAMPI to iteratively estimate GM parameters (λ, a, v), improving reconstruction accuracy.
• A phase-shifter-reduced selection network is proposed to lower power consumption while maintaining estimation performance.
• The algorithm is validated through iterative updates of posterior probabilities and parameter estimation using EM, with closed-form updates for λ, a, and v.

Experimental results

Research questions

• RQ1 Can the SCAMPI algorithm with GM-EM learning outperform conventional compressive sensing methods like OMP and support detection in beamspace mmWave channel estimation?
• RQ2 How does modeling the channel with a Gaussian Mixture prior improve estimation accuracy compared to a uniform prior?
• RQ3 To what extent can the number of phase shifters be reduced without degrading estimation performance?
• RQ4 Does the image reconstruction perspective of the beamspace channel matrix, treating it as a 2D natural image, enable faster and more accurate recovery?
• RQ5 What is the impact of incorporating smoothness and clustering features of mmWave paths into the estimation framework?

Key findings

• The proposed SCAMPI algorithm with GM-EM learning achieves higher estimation accuracy and faster convergence than OMP and support detection algorithms.
• The use of a Gaussian Mixture prior significantly outperforms the uniform prior in terms of reconstruction accuracy and robustness.
• The algorithm remains robust even when the number of phase shifters is reduced by 10%, demonstrating practical hardware resilience.
• The phase-shifter-reduced selection network effectively reduces system power consumption without sacrificing performance.
• The EM-based parameter learning for the GM model enables adaptive and data-driven estimation, improving convergence and accuracy.
• The beamspace channel matrix's inherent sparsity and smoothness are effectively exploited by treating it as a 2D image, enabling high-performance reconstruction.