A Neural Beam Filter for Real-time Multi-channel Speech Enhancement

TL;DR

This paper introduces a causal neural beam filter that leverages spatial-spectral information for real-time multi-channel speech enhancement, outperforming existing methods on a challenging dataset.

Contribution

It proposes a novel neural beam filtering approach with a two-stage process and residual refinement, enhancing speech quality in noisy multi-channel environments.

Findings

Outperforms state-of-the-art multi-channel methods

Effective at low-frequency interference suppression

Demonstrates real-time capability

Abstract

Most deep learning-based multi-channel speech enhancement methods focus on designing a set of beamforming coefficients to directly filter the low signal-to-noise ratio signals received by microphones, which hinders the performance of these approaches. To handle these problems, this paper designs a causal neural beam filter that fully exploits the spatial-spectral information in the beam domain. Specifically, multiple beams are designed to steer towards all directions using a parameterized super-directive beamformer in the first stage. After that, the neural spatial filter is learned by simultaneously modeling the spatial and spectral discriminability of the speech and the interference, so as to extract the desired speech coarsely in the second stage. Finally, to further suppress the interference components especially at low frequencies, a residual estimation module is adopted to refine the output of the second stage. Experimental results demonstrate that the proposed approach outperforms many state-of-the-art multi-channel methods on the generated multi-channel speech dataset based on the DNS-Challenge dataset.