Phase retrieval from short-time Fourier transform in LCA groups
Natalia Accomazzo, Daniel Carando, Rocio Nores, Victoria Paternostro, Sebastian Velazquez
TL;DR
The paper addresses STFT phase retrieval on locally compact abelian groups by constructing window functions and uniformly discrete sampling sets that enable unique recovery of $f\in L^2(K)$ from phaseless samples $|V_g f|$. It develops two key tools: (i) uniformly discrete uniqueness sets for Paley-Wiener spaces and (ii) probabilistic completeness of translates to guarantee phase retrieval from phaseless STFT measurements, extending known results from $\mathbb{R}^d$ to broad LCA groups. By combining these tools and handling product groups, the authors prove Theorem A, establishing STFT phase retrieval on $L^2(K)$ under structural group conditions, and they explain fundamental obstructions to full $L^2(G)$ phase retrieval. The results justify restricting attention to compactly supported subspaces and provide a framework for practical phase retrieval in a wide class of groups, with potential applications in ptychography and time-frequency analysis.
Abstract
We study the short-time Fourier transform phase retrieval problem in locally compact abelian groups. Using probabilistic methods, we show that for a large class of groups $G$ and compact subsets $K\subseteq G$ there exists a window function and a uniformly discrete set in $G\times \widehat{G}$ allowing phase retrieval in $L^2(K)$. We also study the obstructions for STFT phase retrieval on $L^2(G)$, motivating the restriction to compactly supported function spaces.
