Sampling-Frequency-Independent Audio Source Separation Using Convolution Layer Based on Impulse Invariant Method

TL;DR

This paper introduces a convolution layer based on the impulse invariant method that allows a single deep neural network to perform audio source separation across arbitrary sampling frequencies, including unseen ones.

Contribution

The proposed convolution layer enables DNN-based audio source separation models to operate effectively across varying and unseen sampling frequencies.

Findings

Model works with unseen sampling frequencies.

Enables a single model to handle multiple sampling rates.

Improves versatility of audio source separation models.

Abstract

Audio source separation is often used as preprocessing of various applications, and one of its ultimate goals is to construct a single versatile model capable of dealing with the varieties of audio signals. Since sampling frequency, one of the audio signal varieties, is usually application specific, the preceding audio source separation model should be able to deal with audio signals of all sampling frequencies specified in the target applications. However, conventional models based on deep neural networks (DNNs) are trained only at the sampling frequency specified by the training data, and there are no guarantees that they work with unseen sampling frequencies. In this paper, we propose a convolution layer capable of handling arbitrary sampling frequencies by a single DNN. Through music source separation experiments, we show that the introduction of the proposed layer enables a conventional audio source separation model to consistently work with even unseen sampling frequencies.