Documents
Poster
Poster
Perceptually-motivated environment-specific speech enhancement
- Citation Author(s):
- Submitted by:
- Jiaqi Su
- Last updated:
- 10 May 2019 - 1:40am
- Document Type:
- Poster
- Document Year:
- 2019
- Event:
- Paper Code:
- 4382
- Categories:
- Log in to post comments
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.