DMF-Net: A decoupling-style multi-band fusion model for full-band speech enhancement

TL;DR

This paper introduces DMF-Net, a multi-band fusion model that decomposes full-band speech into sub-bands for improved enhancement, outperforming previous methods in speech quality and intelligibility.

Contribution

The paper proposes a decoupling-style multi-band fusion model that estimates full-band speech by separately processing and then fusing sub-band features, enhancing performance over existing approaches.

Findings

Outperforms previous advanced systems in speech quality

Achieves better intelligibility in complex scenarios

Effective multi-stage chain optimization for full-band speech

Abstract

For the difficulty and large computational complexity of modeling more frequency bands, full-band speech enhancement based on deep neural networks is still challenging. Previous studies usually adopt compressed full-band speech features in Bark and ERB scale with relatively low frequency resolution, leading to degraded performance, especially in the high-frequency region. In this paper, we propose a decoupling-style multi-band fusion model to perform full-band speech denoising and dereverberation. Instead of optimizing the full-band speech by a single network structure, we decompose the full-band target into multi sub-band speech features and then employ a multi-stage chain optimization strategy to estimate clean spectrum stage by stage. Specifically, the low- (0-8 kHz), middle- (8-16 kHz), and high-frequency (16-24 kHz) regions are mapped by three separate sub-networks and are then fused to obtain the full-band clean target STFT spectrum. Comprehensive experiments on two public datasets demonstrate that the proposed method outperforms previous advanced systems and yields promising performance in terms of speech quality and intelligibility in real complex scenarios.