Perceptually-motivated Environment-specific Speech Enhancement

Jiaqi Su, Adam Finkelstein, Zeyu JinICASSP 2019 · OpenReview

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.
BibTeX
@inproceedings{su2019perceptually,
  author = {Jiaqi Su and Adam Finkelstein and Zeyu Jin},
  title = {{Perceptually-motivated Environment-specific Speech Enhancement}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2019},
  url = {https://openreview.net/forum?id=7G15ks8Nl6F},
}

Loading related content...

Figure