Princeton > CS Dept > PIXL > Publications Local Access 

Perceptually-motivated Environment-specific Speech Enhancement
Currently in review, January 2019

Jiaqi Su, Adam Finkelstein, Zeyu Jin


Abstract

We introduce a data-driven method to enhance speech recordings made in a specific environment. The method handles denoising, de-reverberation, and equalization matching due recording non-linearities in a unified framework. It relies on a new perceptual loss function that combines adversarial loss with spectrogram features. We show that the method offers an improvement over state of the art baseline methods in both subjective and objective evaluations.

Citation (BibTeX)

Jiaqi Su, Adam Finkelstein, and Zeyu Jin. Perceptually-motivated Environment-specific Speech Enhancement. Currently in review, January 2019.

Links
  Paper preprint
  Listen to audio clips from our experiments.