Single-channel Speech Dereverberation via Generative Adversarial Training

Chenxing Li, Tieqiang Wang, Shuang Xu and Bo Xu

Abstract:

In this paper, we propose a single-channel speech dereverberation system (DeReGAT) based on convolutional, bidirectional long short-term memory and deep feed-forward neural network (CBLDNN) with generative adversarial training (GAT). In order to obtain better speech quality instead of only minimizing a mean square error (MSE), GAT is employed to make the dereverberated speech indistinguishable form the clean samples. Besides, our system can deal with wide range reverberation and be well adapted to variant environments. The experimental results show that the proposed model outperforms weighted prediction error (WPE) and deep neural network-based systems. In addition, DeReGAT is extended to an online speech dereverberation scenario, which reports comparable performance with the offline case.

---

Cite as: Li, C., Wang, T., Xu, S., Xu, B. (2018) Single-channel Speech Dereverberation via Generative Adversarial Training. Proc. Interspeech 2018, 1309-1313, DOI: 10.21437/Interspeech.2018-1234.

---

BiBTeX Entry:

@inproceedings{Li2018,
author={Chenxing Li and Tieqiang Wang and Shuang Xu and Bo Xu},
title={Single-channel Speech Dereverberation via Generative Adversarial Training},
year=2018,
booktitle={Proc. Interspeech 2018},
pages={1309--1313},
doi={10.21437/Interspeech.2018-1234},
url={http://dx.doi.org/10.21437/Interspeech.2018-1234} }