Differentiable Short-Time Fourier Transform: A Time-Frequency Layer with Learnable Parameters

Maxime Leiber, yosra marnissi, Axel Barrau

Keywords: short-time Fourier transform, differentiable architecture search, learnable STFT parameters, time-frequency layer

TL;DR: Differentiable and learnable STFT parameters, differentiable STFT is a time-frequency layer that integrates neural networks.

Abstract: We present a differentiable version of the short-time Fourier transform (STFT), enabling gradient-based optimization of its parameters. This approach integrates with neural networks, allowing joint learning of both STFT and network parameters. Tests on simulated and real data demonstrate an improved time-frequency representation and enhanced performance on downstream tasks, illustrating the potential of our method as a standard for setting spectrogram parameters automatically.

Submission Number: 2