Improved Normalizing Flow-Based Speech Enhancement using an All-pole Gammatone Filterbank for Conditional Input Representation

TL;DR

This paper introduces an improved normalizing flow model for speech enhancement that utilizes an all-pole Gammatone filterbank for conditional input, demonstrating superior perceptual quality over GANs especially at low SNRs.

Contribution

The paper proposes architectural improvements and a novel APG filterbank for NF-based speech enhancement, showing perceptual advantages over GANs.

Findings

NF with APG outperforms GANs in perceptual quality at low SNRs

APG provides high temporal resolution for better speech enhancement

NF approach achieves good quality ratings in listening tests

Abstract

Deep generative models for Speech Enhancement (SE) received increasing attention in recent years. The most prominent example are Generative Adversarial Networks (GANs), while normalizing flows (NF) received less attention despite their potential. Building on previous work, architectural modifications are proposed, along with an investigation of different conditional input representations. Despite being a common choice in related works, Mel-spectrograms demonstrate to be inadequate for the given scenario. Alternatively, a novel All-Pole Gammatone filterbank (APG) with high temporal resolution is proposed. Although computational evaluation metric results would suggest that state-of-the-art GAN-based methods perform best, a perceptual evaluation via a listening test indicates that the presented NF approach (based on time domain and APG) performs best, especially at lower SNRs. On average, APG outputs are rated as having good quality, which is unmatched by the other methods, including GAN.