Compressive Sensing with Prior Support Quality Information and Application to Massive MIMO Channel Estimation with Temporal Correlation

TL;DR

This paper introduces adaptive compressive sensing algorithms that utilize prior support quality information to improve signal recovery, with applications to massive MIMO channel estimation under temporal correlation.

Contribution

It proposes novel CS recovery algorithms that adaptively exploit prior support quality, analyzes their distortion bounds, and demonstrates their effectiveness in massive MIMO channel estimation.

Findings

Better prior support quality improves recovery performance

The proposed algorithm converges in O(log SNR) steps

Robust designs handle incorrect prior support information

Abstract

In this paper, we consider the problem of compressive sensing (CS) recovery with a prior support and the prior support quality information available. Different from classical works which exploit prior support blindly, we shall propose novel CS recovery algorithms to exploit the prior support adaptively based on the quality information. We analyze the distortion bound of the recovered signal from the proposed algorithm and we show that a better quality prior support can lead to better CS recovery performance. We also show that the proposed algorithm would converge in $\mathcal{O}\left(\log\mbox{SNR}\right)$ steps. To tolerate possible model mismatch, we further propose some robustness designs to combat incorrect prior support quality information. Finally, we apply the proposed framework to sparse channel estimation in massive MIMO systems with temporal correlation to further reduce the required pilot training overhead.