Sampling-Frequency-Independent Universal Sound Separation

TL;DR

This paper introduces a sampling-frequency-independent universal sound separation method that maintains performance across various sampling frequencies, addressing a key challenge in creating versatile source separation systems.

Contribution

The paper presents an SF-independent extension of the SuDoRM-RF network using SFI convolutional layers, enabling consistent sound separation across diverse sampling frequencies.

Findings

The proposed method outperforms resampling-based approaches across different SFs.

Signal resampling can degrade separation performance, highlighting the importance of SF-independent methods.

The SFI extension maintains high separation quality without the need for resampling.

Abstract

This paper proposes a universal sound separation (USS) method capable of handling untrained sampling frequencies (SFs). The USS aims at separating arbitrary sources of different types and can be the key technique to realize a source separator that can be universally used as a preprocessor for any downstream tasks. To realize a universal source separator, there are two essential properties: universalities with respect to source types and recording conditions. The former property has been studied in the USS literature, which has greatly increased the number of source types that can be handled by a single neural network. However, the latter property (e.g., SF) has received less attention despite its necessity. Since the SF varies widely depending on the downstream tasks, the universal source separator must handle a wide variety of SFs. In this paper, to encompass the two properties, we propose an SF-independent (SFI) extension of a computationally efficient USS network, SuDoRM-RF. The proposed network uses our previously proposed SFI convolutional layers, which can handle various SFs by generating convolutional kernels in accordance with an input SF. Experiments show that signal resampling can degrade the USS performance and the proposed method works more consistently than signal-resampling-based methods for various SFs.