Consistent Spectrogram Separation from Nonstationary Mixture

TL;DR

This paper introduces a novel spectrogram separation method for nonstationary mixtures, utilizing an inverse problem approach and alternating optimization to produce consistent spectrogram estimates, validated on synthetic and bioacoustic data.

Contribution

It proposes a new inverse problem formulation and an alternating optimization algorithm for consistent spectrogram separation of nonstationary signals.

Findings

Effective separation demonstrated on synthetic mixtures

Algorithm ensures spectrogram consistency

Applied successfully to bioacoustic signals

Abstract

We present a spectrogram separation method tailored for mixtures comprising two nonstationary components. By exploiting the unique characteristics of their time-frequency representations, we propose an inverse problem formulation to estimate the spectrograms of the components. We then introduce an alternating optimization algorithm that ensures the consistency of the estimated spectrograms. The efficacy of the algorithm is evaluated through testing on synthetic mixtures and is applied to a bioacoustic signal.