A new speech enhancement approach based on progressive deep neural networks

Xiaofeng Shu, Yi Zhou, Yin Cao

Research output: Chapter in Book or Report/Conference proceedingConference Proceedingpeer-review

1 Citation (Scopus)

Abstract

In this paper, a speech enhancement method based on the framework of progressive deep neural networks (PDNNs) is proposed to alleviate the recognition performance degradation of the automatic speech recognition (ASR) system in low signal-to-noise ratio (SNR) environments. It aims at decomposing a regression task into multiple subtasks, which are closely related to each other, to improve the system performance. Then the learning targets of these subtasks are designed with gradually increasing SNR gains. Furthermore, a post-processing module, which benefits from the rich information of the learning targets, is applied to further improve the system performance. Experimental results reveal that the proposed method can achieve improvements in both objective and subjective evaluations in low SNR environments when compared with the conventional deep neural network method.

Original languageEnglish
Title of host publication16th International Workshop on Acoustic Signal Enhancement, IWAENC 2018 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages191-195
Number of pages5
ISBN (Electronic)9781538681510
DOIs
Publication statusPublished - 2 Nov 2018
Externally publishedYes
Event16th International Workshop on Acoustic Signal Enhancement, IWAENC 2018 - Tokyo, Japan
Duration: 17 Sept 201820 Sept 2018

Publication series

Name16th International Workshop on Acoustic Signal Enhancement, IWAENC 2018 - Proceedings

Conference

Conference16th International Workshop on Acoustic Signal Enhancement, IWAENC 2018
Country/TerritoryJapan
CityTokyo
Period17/09/1820/09/18

Keywords

  • Learning targets
  • PDNNs
  • SNR
  • Speech enhancement

Fingerprint

Dive into the research topics of 'A new speech enhancement approach based on progressive deep neural networks'. Together they form a unique fingerprint.

Cite this