Determined Audio Source Separation with Multichannel Star Generative Adversarial Network

This paper proposes a multichannel source separation approach, which uses a star generative adversarial network (StarGAN) to model power spectrograms of sources. Various studies have shown the significant contributions of a precise source model to the performance improvement in audio source separation, which indicates the importance of developing a better source model. In this paper, we explore the potential of StarGAN for modeling source spectrograms and investigate the effectiveness of the StarGAN source model in determined multichannel source separation by incorporating it into a frequency-domain independent component analysis (ICA) framework. The experimental results reveal that the proposed StarGAN-based method outperformed conventional methods that use non-negative matrix factorization (NMF) or a variational autoencoder (VAE) for source spectrogram modeling.