AudioSlots: A slot-centric generative model for audio separation
This work addresses audio source separation for applications like speech processing, but it is incremental as it adapts existing vision techniques to audio.
The authors tackled blind source separation in audio by introducing AudioSlots, a slot-centric generative model inspired by object-centric vision methods, achieving a proof of concept on Libri2Mix speech separation.
In a range of recent works, object-centric architectures have been shown to be suitable for unsupervised scene decomposition in the vision domain. Inspired by these methods we present AudioSlots, a slot-centric generative model for blind source separation in the audio domain. AudioSlots is built using permutation-equivariant encoder and decoder networks. The encoder network based on the Transformer architecture learns to map a mixed audio spectrogram to an unordered set of independent source embeddings. The spatial broadcast decoder network learns to generate the source spectrograms from the source embeddings. We train the model in an end-to-end manner using a permutation invariant loss function. Our results on Libri2Mix speech separation constitute a proof of concept that this approach shows promise. We discuss the results and limitations of our approach in detail, and further outline potential ways to overcome the limitations and directions for future work.