Zobrazit v EDS

Speech Perception Improvement Algorithm Based on a Dual-Path Long Short-Term Memory Network

Uloženo v:
Podrobná bibliografie
Název: Speech Perception Improvement Algorithm Based on a Dual-Path Long Short-Term Memory Network
Autoři: Hyeong Il Koh, Sungdae Na, Myoung Nam Kim
Zdroj: Bioengineering, Vol 10, Iss 11, p 1325 (2023)
Informace o vydavateli: MDPI AG, 2023.
Rok vydání: 2023
Sbírka: LCC:Technology
LCC:Biology (General)
Témata: speech enhancement, STFT, LSTM, encoder–decoder structure, dual-path network, spectral extension block, Technology, Biology (General), QH301-705.5
Popis: Current deep learning-based speech enhancement methods focus on enhancing the time–frequency representation of the signal. However, conventional methods can lead to speech damage due to resolution mismatch problems that emphasize only specific information in the time or frequency domain. To address these challenges, this paper introduces a speech enhancement model designed with a dual-path structure that identifies key speech characteristics in both the time and time–frequency domains. Specifically, the time path aims to model semantic features hidden in the waveform, while the time–frequency path attempts to compensate for the spectral details via a spectral extension block. These two paths enhance temporal and spectral features via mask functions modeled as LSTM, respectively, offering a comprehensive approach to speech enhancement. Experimental results show that the proposed dual-path LSTM network consistently outperforms conventional single-domain speech enhancement methods in terms of speech quality and intelligibility.
Druh dokumentu: article
Popis souboru: electronic resource
Jazyk: English
ISSN: 2306-5354
Relation: https://www.mdpi.com/2306-5354/10/11/1325; https://doaj.org/toc/2306-5354
DOI: 10.3390/bioengineering10111325
Přístupová URL adresa: https://doaj.org/article/29ba1fb2bebd49e4b183fd9e1f8573bd
Přístupové číslo: edsdoj.29ba1fb2bebd49e4b183fd9e1f8573bd
Databáze: Directory of Open Access Journals
Popis
Abstrakt:Current deep learning-based speech enhancement methods focus on enhancing the time–frequency representation of the signal. However, conventional methods can lead to speech damage due to resolution mismatch problems that emphasize only specific information in the time or frequency domain. To address these challenges, this paper introduces a speech enhancement model designed with a dual-path structure that identifies key speech characteristics in both the time and time–frequency domains. Specifically, the time path aims to model semantic features hidden in the waveform, while the time–frequency path attempts to compensate for the spectral details via a spectral extension block. These two paths enhance temporal and spectral features via mask functions modeled as LSTM, respectively, offering a comprehensive approach to speech enhancement. Experimental results show that the proposed dual-path LSTM network consistently outperforms conventional single-domain speech enhancement methods in terms of speech quality and intelligibility.
ISSN:23065354
DOI:10.3390/bioengineering10111325