Well-Posedness of Sparse Frequency Estimation

Abstract

The problem of estimating the frequencies of an exponential sum has been studied extensively over the last years. It can be understood as a sparse estimation problem, as it strives to identify the sparse representation of a signal using exponentials. In this paper, we are interested in its intrinsic stability properties. We derive a bound very similar to the restricted isometry property. Conditional well-posedness follows: Any exponential sum with samples close to the unknown ground truth has close frequencies as well, provided that it satisfies our model assumptions. The most important assumption is that the frequencies are well-separated. Furthermore, we show that the presented bound is sharp and gives rise to improved estimates of condition numbers of certain Vandermonde matrices.