Sudo rm -rf: Efficient Networks for Universal Audio Source Separation

TL;DR

This paper introduces an efficient neural network architecture for universal audio source separation that achieves high-quality results with fewer computational resources, outperforming some state-of-the-art methods.

Contribution

The paper proposes the SuDoRMRF architecture, a novel convolutional network that efficiently combines multi-resolution features for audio separation, reducing resource needs while maintaining or improving performance.

Findings

SuDoRMRF achieves comparable or better separation quality than state-of-the-art methods.

The model requires fewer floating point operations and less memory.

It performs well on both speech and environmental sound datasets.

Abstract

In this paper, we present an efficient neural network for end-to-end general purpose audio source separation. Specifically, the backbone structure of this convolutional network is the SUccessive DOwnsampling and Resampling of Multi-Resolution Features (SuDoRMRF) as well as their aggregation which is performed through simple one-dimensional convolutions. In this way, we are able to obtain high quality audio source separation with limited number of floating point operations, memory requirements, number of parameters and latency. Our experiments on both speech and environmental sound separation datasets show that SuDoRMRF performs comparably and even surpasses various state-of-the-art approaches with significantly higher computational resource requirements.