Official repository for LMFCA-Net: A Lightweight Model for Multi-Channel Speech Enhancement with Efficient Narrow-Band and Cross-Band Attention. This work has been accepted at ICASSP 2025.
The training scripts for LMFCA-Net are adapted from McNet: Fuse Multiple Cues for Multichannel Speech Enhancement. We extend our gratitude to them for their valuable contributions.