Differentiable Time-Frequency Scattering on GPU

TL;DR

This paper introduces a GPU-compatible, differentiable implementation of joint time-frequency scattering (JTFS) in Python, enabling efficient analysis and application in audio perception and synthesis tasks.

Contribution

It provides a flexible, portable JTFS implementation supporting multiple backends, addressing previous limitations in differentiability, speed, and flexibility.

Findings

Enables unsupervised manifold learning of spectrotemporal modulations

Improves supervised classification of musical instruments

Facilitates texture resynthesis of bioacoustic sounds

Abstract

Joint time-frequency scattering (JTFS) is a convolutional operator in the time-frequency domain which extracts spectrotemporal modulations at various rates and scales. It offers an idealized model of spectrotemporal receptive fields (STRF) in the primary auditory cortex, and thus may serve as a biological plausible surrogate for human perceptual judgments at the scale of isolated audio events. Yet, prior implementations of JTFS and STRF have remained outside of the standard toolkit of perceptual similarity measures and evaluation methods for audio generation. We trace this issue down to three limitations: differentiability, speed, and flexibility. In this paper, we present an implementation of time-frequency scattering in Python. Unlike prior implementations, ours accommodates NumPy, PyTorch, and TensorFlow as backends and is thus portable on both CPU and GPU. We demonstrate the usefulness of JTFS via three applications: unsupervised manifold learning of spectrotemporal modulations, supervised classification of musical instruments, and texture resynthesis of bioacoustic sounds.