Skip to main content

Perceptually-motivated Environment-specific Speech Enhancement

Author(s): Su, Jiaqi; Finkelstein, Adam; Jin, Zeyu

To refer to this page use:
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.
Publication Date: 2019
Citation: Su, Jiaqi, Adam Finkelstein, and Zeyu Jin. "Perceptually-motivated Environment-specific Speech Enhancement." IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2019): pp. 7015-7019. IEEE, 2019. doi:10.1109/ICASSP.2019.8683654
DOI: 10.1109/ICASSP.2019.8683654
ISSN: 1520-6149
EISSN: 2379-190X
Pages: 7015 - 7019
Type of Material: Conference Article
Journal/Proceeding Title: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Version: Author's manuscript

Items in OAR@Princeton are protected by copyright, with all rights reserved, unless otherwise indicated.