SDNet: Noise-Robust Bandwidth Extension under Flexible Sampling Rates

Abstract

Bandwidth extension (BWE), also known as audio super-resolution (SR), aims to predict a high resolution (HR) speech signal from its low resolution (LR) corresponding part. Most neural BWE models work at a specific sampling rate but, producing the final result in a noise-free environment by recovering the spectrogram of high-frequency part of the signal and concatenating it with the original low-frequency part. Although these methods achieve high accuracy, they become less effective when facing the real-world scenario, where unavoidable noise is present and sampling rates are flexible. To address this problem, we propose Super Denoise Net (SDNet), a neural network for a joint task of BWE and noise reduction from a flexible low sampling rate signal. To that end, we design gated convolution and lattice convolution blocks to enhance the repair capability and capture information in the time-frequency axis, respectively. The experiments show our method outperforms all current state-of-the-art (SOTA) noise-robust BWE model in Valentini-Botinhao test set. Our model also outperforms other baselines on DNS 2020 no-reverb test set with higher objective and subjective scores.