Audio Separation Method Based on ECA Attention Mechanism

In our daily communication environment, speech transmission is usually disturbed by various noises. Although humans can easily focus their attention on the desired sound, establishing an effective system to address this issue presents significant challenges. Current research in this area is not yet deeply explored, resulting in some errors in experimental results. To solve this problem, this paper further improves on U-Net, and proposes a audio separation model ECAU-Net based on ECA attention mechanism. By incorporating the ECA mechanism between the downsampling blocks of the model's encoder, the accuracy and quality of audio separation are improved. The experimental results show that the improved audio separation method has achieved good results.