Raw Multi-Channel Audio Source Separation using Multi-Resolution Convolutional Auto-Encoders

TL;DR

This paper presents a novel multi-channel convolutional auto-encoder that operates directly on raw audio signals to effectively separate singing voice from stereo music without relying on handcrafted features.

Contribution

It introduces a multi-resolution auto-encoder architecture that learns to extract features directly from raw signals for source separation, eliminating the need for feature engineering.

Findings

Achieves effective multi-channel source separation from raw signals.

No need for handcrafted features or pre/post-processing.

Demonstrates competitive performance on stereo music separation.

Abstract

Supervised multi-channel audio source separation requires extracting useful spectral, temporal, and spatial features from the mixed signals. The success of many existing systems is therefore largely dependent on the choice of features used for training. In this work, we introduce a novel multi-channel, multi-resolution convolutional auto-encoder neural network that works on raw time-domain signals to determine appropriate multi-resolution features for separating the singing-voice from stereo music. Our experimental results show that the proposed method can achieve multi-channel audio source separation without the need for hand-crafted features or any pre- or post-processing.