LMFCA-Net: A Lightweight Model for Multi-Channel Speech Enhancement with Efficient Narrow-Band and Cross-Band Attention

TL;DR

LMFCA-Net is a lightweight multi-channel speech enhancement model that uses decoupled attention mechanisms to efficiently capture spectral and cross-band information, achieving high performance with lower computational cost.

Contribution

It introduces T-FCA and F-FCA mechanisms for efficient long-range feature capture without recurrent units, advancing practical speech enhancement solutions.

Findings

Performs comparably to state-of-the-art methods

Reduces computational complexity and latency

Suitable for real-time applications

Abstract

Deep learning based end-to-end multi-channel speech enhancement methods have achieved impressive performance by leveraging sub-band, cross-band, and spatial information. However, these methods often demand substantial computational resources, limiting their practicality on terminal devices. This paper presents a lightweight multi-channel speech enhancement network with decoupled fully connected attention (LMFCA-Net). The proposed LMFCA-Net introduces time-axis decoupled fully-connected attention (T-FCA) and frequency-axis decoupled fully-connected attention (F-FCA) mechanisms to effectively capture long-range narrow-band and cross-band information without recurrent units. Experimental results show that LMFCA-Net performs comparably to state-of-the-art methods while significantly reducing computational complexity and latency, making it a promising solution for practical applications.