Channel and temporal-frequency attention UNet for monaural speech enhancement

UNet с канальным и временно-частотным вниманием для улучшения монофонической речи
Mingjiang Wang, Shiyun Xu, Zehua Zhang
2023-08-14

UNetchannel and temporal-frequency attentioncomplex ideal ratio maskdeep noise suppressionmonaural speech enhancement
Abstract The presence of noise and reverberation significantly impedes speech clarity and intelligibility. To mitigate these effects, numerous deep learning-based network models have been proposed for speech enhancement tasks aimed at improving speech quality. In this study, we propose a monaural speech enhancement model called the channel and temporal-frequency attention UNet (CTFUNet). CTFUNet takes the noisy spectrum as input and produces a complex ideal ratio mask (cIRM) as output. To improve the speech enhancement performance of CTFUNet, we employ multi-scale temporal-frequency processing to extract input speech spectrum features. We also utilize multi-conv head channel attention and residual channel attention to capture temporal-frequency and channel features. Moreover, we introduce the channel temporal-frequency skip connection to alleviate information loss between down-sampling and up-sampling. On the blind test set of the first deep noise suppression challenge, our proposed CTFUNet has better denoising performance than the champion models and the latest models. Furthermore, our model outperforms recent models such as Uformar and MTFAA in both denoising and dereverberation performance.
1
A channel temporal-frequency skip connection is introduced to reduce information loss between UNet down-sampling and up-sampling stages.
2
CTFUNet is a monaural speech-enhancement model that maps noisy spectra to complex ideal ratio masks (cIRMs).
3
CTFUNet outperforms Uformer and MTFAA in both denoising and dereverberation performance.
4
On the first Deep Noise Suppression Challenge blind test set, CTFUNet achieves better denoising performance than both champion and recent competing models.
5
The model uses multi-scale temporal-frequency processing, multi-convolution-head channel attention, and residual channel attention to extract speech and channel features.

monaural noisy and reverberant speech signals

speech denoising and dereverberation performance, including speech quality and intelligibility improvement

Publication Details
Publication Date
2023-08-14
Journal
Publisher
ISSN
Cited by
21
Access Type
Author Information
Authors
Mingjiang Wang
Shiyun Xu
Zehua Zhang
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%