Deep Denoising Autoencoder Based Post Filtering for Speech Enhancement

Ryandhimas E. Zezario, Jen Wei Huang, Xugang Lu, Yu Tsao, Hsin Te Hwang, Hsin Min Wang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Citations (Scopus)

Abstract

In this paper, we present a simple yet effective deep denoising autoencoder (DDAE) based post-filter (DPF) approach for speech enhancement (SE). The DPF is designed to estimate the spectral difference of clean-noisy speech pair based on the enhanced-noisy speech pair. The difference estimated by the DPF approach is then used to compensate the noisy speech to obtain the final enhanced speech. We integrate the proposed DPF approach with one traditional SE method (minimum mean square error) and one deep-learning-based SE method (DDAE). Experiments on various noise types and signal-to-noise-ratio conditions were carried out to test the integrated systems. Results of three standardized objective evaluation metrics and automatic speech recognition (ASR) tests confirm that integrating the proposed DPF can improve the performance in further reducing spectral distortions and enhancing the speech quality and intelligibility.

Original languageEnglish
Title of host publication2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages373-377
Number of pages5
ISBN (Electronic)9789881476852
DOIs
Publication statusPublished - 2018 Jul 2
Event10th Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Honolulu, United States
Duration: 2018 Nov 122018 Nov 15

Publication series

Name2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018 - Proceedings

Conference

Conference10th Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2018
Country/TerritoryUnited States
CityHonolulu
Period18-11-1218-11-15

All Science Journal Classification (ASJC) codes

  • Information Systems

Fingerprint

Dive into the research topics of 'Deep Denoising Autoencoder Based Post Filtering for Speech Enhancement'. Together they form a unique fingerprint.

Cite this