Exact reconstruction of sparse non-harmonic signals from their Fourier coefficients (Q2059803)

In this paper, the authors propose a new method to reconstruct real non-harmonic Fourier sums, i.e. real signals which can be represented as sparse exponential sums of the form \[ f(t)=\sum_{j=1}^{K}\gamma_{j}\cos(2\pi a_j t + b_j), \] from their Fourier coefficients. They assume that \(K\in\mathbb{N}\), \(\gamma_j\in (0,\infty)\), \((a_j, b_j) \in (0, \infty)\times[0, 2\pi )\), and that the frequency parameters \(a_j\) are pairwise distinct. Their approach is based on two steps. The first one consists of reconstructing the non-\(P\)-periodic part of \(f\), employing a modification of the recently proposed AAA algorithm [\textit{Y. Nakatsukasa} et al., SIAM J. Sci. Comput. 40, No. 3, A1494--A1522 (2018; Zbl 1390.41015)]; the second step concerns the determination of possible \(P\)-periodic terms of \(f\). In particular, they prove that their method allows to uniquely determine \(f\) from at most \(2K+2\) of its Fourier coefficients. Finally, the authors present two numerical experiments, which show that the considered reconstruction scheme provides very good reconstruction results even for small frequency gaps if \(P\) is chosen suitably. These results are compared with the ones obtained with a stabilized variant of Prony's method.