Advanced search
1 file | 741.25 KB Add to list

SERGAN : speech enhancement using relativistic generative adversarial networks with gradient penalty

Deepak Baby (UGent) and Sarah Verhulst (UGent)
Author
Organization
Abstract
Popular neural network-based speech enhancement systems operate on the magnitude spectrogram and ignore the phase mismatch between the noisy and clean speech signals. Recently, conditional generative adversarial networks (cGANs) have shown promise in addressing the phase mismatch problem by directly mapping the raw noisy speech waveform to the underlying clean speech signal. However, stabilizing and training cGAN systems is difficult and they still fall short of the performance achieved by spectral enhancement approaches. This paper introduces relativistic GANs with a relativistic cost function at its discriminator and gradient penalty to improve time-domain speech enhancement. Simulation results show that relativistic discriminators provide a more stable training of cGANs and yield a better generator network for improved speech enhancement performance.
Keywords
speech enhancement, relativistic GAN, convolutional neural networks

Downloads

  • (...).pdf
    • full text
    • |
    • UGent only
    • |
    • PDF
    • |
    • 741.25 KB

Citation

Please use this url to cite or link to this publication:

MLA
Baby, Deepak, and Sarah Verhulst. SERGAN : Speech Enhancement Using Relativistic Generative Adversarial Networks with Gradient Penalty. 2019, pp. 106–10.
APA
Baby, D., & Verhulst, S. (2019). SERGAN : speech enhancement using relativistic generative adversarial networks with gradient penalty (pp. 106–110). Presented at the 44th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Brighton, ENGLAND.
Chicago author-date
Baby, Deepak, and Sarah Verhulst. 2019. “SERGAN : Speech Enhancement Using Relativistic Generative Adversarial Networks with Gradient Penalty.” In , 106–10.
Chicago author-date (all authors)
Baby, Deepak, and Sarah Verhulst. 2019. “SERGAN : Speech Enhancement Using Relativistic Generative Adversarial Networks with Gradient Penalty.” In , 106–110.
Vancouver
1.
Baby D, Verhulst S. SERGAN : speech enhancement using relativistic generative adversarial networks with gradient penalty. In 2019. p. 106–10.
IEEE
[1]
D. Baby and S. Verhulst, “SERGAN : speech enhancement using relativistic generative adversarial networks with gradient penalty,” presented at the 44th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Brighton, ENGLAND, 2019, pp. 106–110.
@inproceedings{8613639,
  abstract     = {Popular neural network-based speech enhancement systems operate on the magnitude spectrogram and ignore the phase mismatch between the noisy and clean speech signals. Recently, conditional generative adversarial networks (cGANs) have shown promise in addressing the phase mismatch problem by directly mapping the raw noisy speech waveform to the underlying clean speech signal. However, stabilizing and training cGAN systems is difficult and they still fall short of the performance achieved by spectral enhancement approaches. This paper introduces relativistic GANs with a relativistic cost function at its discriminator and gradient penalty to improve time-domain speech enhancement. Simulation results show that relativistic discriminators provide a more stable training of cGANs and yield a better generator network for improved speech enhancement performance.},
  author       = {Baby, Deepak and Verhulst, Sarah},
  isbn         = {9781538646588},
  issn         = {1520-6149},
  keywords     = {speech enhancement,relativistic GAN,convolutional neural networks},
  language     = {eng},
  location     = {Brighton, ENGLAND},
  pages        = {106--110},
  title        = {SERGAN : speech enhancement using relativistic generative adversarial networks with gradient penalty},
  url          = {http://dx.doi.org/10.1109/ICASSP.2019.8683799},
  year         = {2019},
}

Altmetric
View in Altmetric
Web of Science
Times cited: