Phase retrieval with Bregman divergences: Application to audio signal recovery

TL;DR

This paper introduces a novel phase retrieval method using Bregman divergences tailored for audio signals, along with an efficient gradient algorithm for its implementation.

Contribution

It presents a new formulation of phase retrieval based on Bregman divergences and develops a fast gradient algorithm for audio signal recovery.

Findings

Effective phase retrieval using Bregman divergences for audio signals

Development of a fast gradient algorithm for the proposed method

Potential improvements over quadratic cost-based approaches

Abstract

Phase retrieval aims to recover a signal from magnitude or power spectra measurements. It is often addressed by considering a minimization problem involving a quadratic cost function. We propose a different formulation based on Bregman divergences, which encompass divergences that are appropriate for audio signal processing applications. We derive a fast gradient algorithm to solve this problem.