Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

Perceptually-motivated environment-specific speech enhancement

Abstract: 

This paper introduces a deep learning approach to enhance speech recordings made in a specific environment. A single neural network learns to ameliorate several types of recording artifacts, including noise, reverberation, and non-linear equalization. The method relies on a new perceptual loss function that combines adversarial loss with spectrogram features. Both subjective and objective evaluations show that the proposed approach improves on state-of-the-art baseline methods.

up
0 users have voted:

Paper Details

Authors:
Jiaqi Su, Adam Finkelstein, Zeyu Jin
Submitted On:
10 May 2019 - 1:40am
Short Link:
Type:
Poster
Event:
Paper Code:
4382
Document Year:
2019
Cite

Document Files

ICASSP2019_SU_SE_poster

(68)

ICASSP2019_SU_SE_poster

(70)

Subscribe

[1] Jiaqi Su, Adam Finkelstein, Zeyu Jin, "Perceptually-motivated environment-specific speech enhancement", IEEE SigPort, 2019. [Online]. Available: http://sigport.org/4272. Accessed: Jul. 13, 2020.
@article{4272-19,
url = {http://sigport.org/4272},
author = {Jiaqi Su; Adam Finkelstein; Zeyu Jin },
publisher = {IEEE SigPort},
title = {Perceptually-motivated environment-specific speech enhancement},
year = {2019} }
TY - EJOUR
T1 - Perceptually-motivated environment-specific speech enhancement
AU - Jiaqi Su; Adam Finkelstein; Zeyu Jin
PY - 2019
PB - IEEE SigPort
UR - http://sigport.org/4272
ER -
Jiaqi Su, Adam Finkelstein, Zeyu Jin. (2019). Perceptually-motivated environment-specific speech enhancement. IEEE SigPort. http://sigport.org/4272
Jiaqi Su, Adam Finkelstein, Zeyu Jin, 2019. Perceptually-motivated environment-specific speech enhancement. Available at: http://sigport.org/4272.
Jiaqi Su, Adam Finkelstein, Zeyu Jin. (2019). "Perceptually-motivated environment-specific speech enhancement." Web.
1. Jiaqi Su, Adam Finkelstein, Zeyu Jin. Perceptually-motivated environment-specific speech enhancement [Internet]. IEEE SigPort; 2019. Available from : http://sigport.org/4272