Sergan: Speech Enhancement Using Relativistic Generative Adversarial Networks with Gradient Penalty
By Deepak Baby, Sarah Verhulst. 2019
121 citations recorded by the source corpus.
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.
Subfields: Computational Mechanics, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computational Mechanics
In Signal Processing
- Yossi Adi
- Gabriel Synnaeve
- Alexandre D'efossez
- Szu-Wei Fu
- Yu Tsao
- Chien-Feng Liao
- Shou-De Lin
- Wei-Ning Hsu
Citations in the corpus
- Real Time Speech Enhancement in the Waveform Domain (2020, background): “Over the last years, there has been a growing interest towards DNN based methods for speech enhancement [34, 35, 36, 37, 20, 38, 39, 22, 40, 7, 41, 8, 42, 21, 37].”
- MetricGAN: Generative Adversarial Networks based Black-box Metric Scores Optimization for Speech Enhancement (2019, methodology): “Similar to the structure of SEGAN , WGAN-GP and SERGAN (Baby & Verhulst, 2019) introduced Wasserstein loss and relativistic least-square loss for GAN training, respectively.”