A general framework for online audio source separation

TL;DR

This paper introduces a comprehensive online audio source separation framework that integrates spatial and spectral cues, utilizing a GEM algorithm for parameter estimation, and compares its performance to offline methods.

Contribution

It proposes a novel general framework combining spatial and spectral cues for online audio source separation, using ML estimation with a GEM algorithm.

Findings

Performance varies with block and step sizes.

The framework approaches offline accuracy levels.

It outperforms existing online methods in certain conditions.

Abstract

We consider the problem of online audio source separation. Existing algorithms adopt either a sliding block approach or a stochastic gradient approach, which is faster but less accurate. Also, they rely either on spatial cues or on spectral cues and cannot separate certain mixtures. In this paper, we design a general online audio source separation framework that combines both approaches and both types of cues. The model parameters are estimated in the Maximum Likelihood (ML) sense using a Generalised Expectation Maximisation (GEM) algorithm with multiplicative updates. The separation performance is evaluated as a function of the block size and the step size and compared to that of an offline algorithm.