Residual Unet with Attention Mechanism for Time-Frequency Domain Speech Enhancement

• DOI: 10.23919/ccc55666.2022.9902215
• Publication Date: 2022-10-11T15:33:35Z
• Authors: Hanyu Chen; Xiwei Peng; Qiqi Jiang; Yujie Guo
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Eliminating the negative effects of background environmental noise is an interesting and challenging task in audio processing. In recent years, denoising technology based on neural networks (NN) has achieved good performance. particular, structure convolutional encoder decoder been proven to achieve enhancement effects. On this basis, paper proposes a residual unet combined with attention mechanism. Effectively reduce impact gradient disappearance network training, improve semantic gap between output due shortcut connections. The experimental results show that compared DNN baseline network, enhanced voice quality significantly improved.