A two-step backward compatible fullband speech enhancement system

TL;DR

This paper introduces a two-step fullband speech enhancement system that maintains backward compatibility with existing wideband systems while achieving high-quality enhancement at 48kHz sample rate.

Contribution

It presents a novel two-step approach for fullband speech enhancement that ensures backward compatibility with wideband systems, unlike existing single-network fullband methods.

Findings

Achieves high-quality fullband speech enhancement at 48kHz

Ensures backward compatibility with wideband systems

Outperforms existing single-network fullband methods

Abstract

Speech enhancement methods based on deep learning have surpassed traditional methods. While many of these new approaches are operating on the wideband (16kHz) sample rate, a new fullband (48kHz) speech enhancement system is proposed in this paper. Compared to the existing fullband systems that utilizes perceptually motivated features to train the fullband speech enhancement using a single network structure, the proposed system is a two-step system ensuring good fullband speech enhancement quality while backward compatible to the existing wideband systems.