Parallel gated neural network with attention mechanism for speech enhancement
Parallel gated neural network with attention mechanism for speech enhancement
Deep learning algorithm are increasingly used for speech enhancement (SE). In supervised methods, global and local information is required for accurate spectral mapping. A key restriction is often poor capture of key contextual information. To leverage long-term for target speakers and compensate distortions of cleaned speech, this paper adopts a sequence-to-sequence (S2S) mapping structure and proposes a novel monaural speech enhancement system, consisting of a Feature Extraction Block (FEB), a Compensation Enhancement Block (ComEB) and a Mask Block (MB). In the FEB a U-net block is used to extract abstract features using complex-valued spectra with one path to suppress the background noise in the magnitude domain using masking methods and the MB takes magnitude features from the FEB and compensates the lost complex-domain features produced from ComEB to restore the final cleaned speech. Experiments are conducted on the Librispeech dataset and results show that the proposed model obtains a good performance in terms of ESTOI and PESQ scores.
complex domain compensation, global and local speech information, magnitude domain mask, sequence-to-sequence mapping, Supervised speech enhancement
197-201
Cui, Jianqiao
3961d0d6-9687-4fbc-9e17-93be8bd86a36
Bleeck, Stefan
c888ccba-e64c-47bf-b8fa-a687e87ec16c
27 September 2023
Cui, Jianqiao
3961d0d6-9687-4fbc-9e17-93be8bd86a36
Bleeck, Stefan
c888ccba-e64c-47bf-b8fa-a687e87ec16c
Cui, Jianqiao and Bleeck, Stefan
(2023)
Parallel gated neural network with attention mechanism for speech enhancement.
In 2023 6th International Conference on Big Data and Artificial Intelligence, BDAI 2023.
IEEE.
.
(doi:10.1109/BDAI59165.2023.10256776).
Record type:
Conference or Workshop Item
(Paper)
Abstract
Deep learning algorithm are increasingly used for speech enhancement (SE). In supervised methods, global and local information is required for accurate spectral mapping. A key restriction is often poor capture of key contextual information. To leverage long-term for target speakers and compensate distortions of cleaned speech, this paper adopts a sequence-to-sequence (S2S) mapping structure and proposes a novel monaural speech enhancement system, consisting of a Feature Extraction Block (FEB), a Compensation Enhancement Block (ComEB) and a Mask Block (MB). In the FEB a U-net block is used to extract abstract features using complex-valued spectra with one path to suppress the background noise in the magnitude domain using masking methods and the MB takes magnitude features from the FEB and compensates the lost complex-domain features produced from ComEB to restore the final cleaned speech. Experiments are conducted on the Librispeech dataset and results show that the proposed model obtains a good performance in terms of ESTOI and PESQ scores.
This record has no associated files available for download.
More information
Published date: 27 September 2023
Venue - Dates:
6th International Conference on Big Data and Artificial Intelligence, BDAI 2023, , Haining, China, 2023-07-07 - 2023-07-09
Keywords:
complex domain compensation, global and local speech information, magnitude domain mask, sequence-to-sequence mapping, Supervised speech enhancement
Identifiers
Local EPrints ID: 492462
URI: http://eprints.soton.ac.uk/id/eprint/492462
PURE UUID: d8a8e4ec-3071-4059-8533-cbaac6d00793
Catalogue record
Date deposited: 29 Jul 2024 16:53
Last modified: 30 Jul 2024 01:57
Export record
Altmetrics
Contributors
Author:
Jianqiao Cui
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics