BSDB-Net: Band-Split Dual-Branch Network with Selective State Spaces Mechanism for Monaural Speech Enhancement

TL;DR

BSDB-Net introduces a dual-branch network that decouples amplitude and phase for monaural speech enhancement, employing band-split and Mamba modules to significantly reduce complexity while maintaining high performance.

Contribution

The paper proposes a novel dual-path network with band-split strategy and Mamba modules, effectively decoupling amplitude and phase and reducing computational complexity in speech enhancement.

Findings

Achieves 8.3x reduction in computational complexity compared to baselines.

Maintains superior speech enhancement performance despite reduced complexity.

Reduces complexity by 25x compared to transformer-based models.

Abstract

Although the complex spectrum-based speech enhancement(SE) methods have achieved significant performance, coupling amplitude and phase can lead to a compensation effect, where amplitude information is sacrificed to compensate for the phase that is harmful to SE. In addition, to further improve the performance of SE, many modules are stacked onto SE, resulting in increased model complexity that limits the application of SE. To address these problems, we proposed a dual-path network based on compressed frequency using Mamba. First, we extract amplitude and phase information through parallel dual branches. This approach leverages structured complex spectra to implicitly capture phase information and solves the compensation effect by decoupling amplitude and phase, and the network incorporates an interaction module to suppress unnecessary parts and recover missing components from the other branch. Second, to reduce network complexity, the network introduces a band-split strategy to compress the frequency dimension. To further reduce complexity while maintaining good performance, we designed a Mamba-based module that models the time and frequency dimensions under linear complexity. Finally, compared to baselines, our model achieves an average 8.3 times reduction in computational complexity while maintaining superior performance. Furthermore, it achieves a 25 times reduction in complexity compared to transformer-based models.