Threshold Effects in Parameter Estimation from Compressed Data

TL;DR

This paper analyzes the threshold effects in parameter estimation from compressed noisy data, focusing on subspace swaps that cause performance degradation below certain SNR levels, and provides bounds and case studies on this phenomenon.

Contribution

It derives analytical bounds on subspace swap probabilities and explores the impact of compression on threshold SNR in parameter estimation.

Findings

Threshold effects cause mean-squared error to deviate from the Cramer-Rao bound below a certain SNR.

Analytical bounds on subspace swap probability are established for compressed noisy data.

Doubling compression ratio increases threshold SNR by approximately 3 dB.

Abstract

In this paper, we investigate threshold effects associated with swapping of signal and noise subspaces in estimating signal parameters from compressed noisy data. The term threshold effect refers to a sharp departure of mean-squared error from the Cramer-Rao bound when the signal-to-noise ratio falls below a threshold SNR. In many cases, the threshold effect is caused by a subspace swap event, when the measured data (or its sample covariance) is better approximated by a subset of components of an orthogonal subspace than by the components of a signal subspace. We derive analytical lower bounds on the probability of a subspace swap in compressively measured noisy data. These bounds guide our understanding of threshold effects and performance breakdown for parameter estimation using compression. As a case study, we investigate threshold effects in maximum likelihood (ML) estimation of directions of arrival of two closely-spaced sources using co-prime subsampling. Our results show the impact of compression on threshold SNR. A rule of thumb is that every doubling of compression ratio brings a penalty in threshold SNR of 3 dB.