Deep Bayesian Unsupervised Source Separation Based on a Complex Gaussian Mixture Model

TL;DR

This paper introduces an unsupervised neural source separation method using a complex Gaussian mixture model that jointly trains separation and localization networks, improving performance without extensive supervised data.

Contribution

It proposes a deep Bayesian framework that jointly trains separation and localization networks using a complex Gaussian mixture model, addressing frequency permutation ambiguity.

Findings

Outperformed conventional initialization methods in simulated speech mixtures

Effectively estimates spatial variables without supervised training data

Enhances multichannel source separation performance

Abstract

This paper presents an unsupervised method that trains neural source separation by using only multichannel mixture signals. Conventional neural separation methods require a lot of supervised data to achieve excellent performance. Although multichannel methods based on spatial information can work without such training data, they are often sensitive to parameter initialization and degraded with the sources located close to each other. The proposed method uses a cost function based on a spatial model called a complex Gaussian mixture model (cGMM). This model has the time-frequency (TF) masks and direction of arrivals (DoAs) of sources as latent variables and is used for training separation and localization networks that respectively estimate these variables. This joint training solves the frequency permutation ambiguity of the spatial model in a unified deep Bayesian framework. In addition, the pre-trained network can be used not only for conducting monaural separation but also for efficiently initializing a multichannel separation algorithm. Experimental results with simulated speech mixtures showed that our method outperformed a conventional initialization method.