# A Style Transfer Approach to Source Separation

## TL;DR

This paper introduces a novel style transfer approach to source separation using cycle-consistent variational auto-encoders, enabling separation without paired training data by learning shared representations across mixture and clean sound domains.

## Contribution

It proposes a new method that treats source separation as a style transfer problem, removing the need for paired training data through shared latent representations.

## Key findings

- Achieves source separation without paired supervision.
- Utilizes cycle-consistent variational auto-encoders for domain mapping.
- Demonstrates effective separation in experimental results.

## Abstract

Training neural networks for source separation involves presenting a mixture recording at the input of the network and updating network parameters in order to produce an output that resembles the clean source. Consequently, supervised source separation depends on the availability of paired mixture-clean training examples. In this paper, we interpret source separation as a style transfer problem. We present a variational auto-encoder network that exploits the commonality across the domain of mixtures and the domain of clean sounds and learns a shared latent representation across the two domains. Using these cycle-consistent variational auto-encoders, we learn a mapping from the mixture domain to the domain of clean sounds and perform source separation without explicitly supervising with paired training examples.