/Two-dimensional-Self-attention-based-Speech-Enhancement

A 2-dimensional Self-attention-based Solution with Cooperative Gated Convolutional Modules for Speech Enhancement

Primary LanguagePython

Two-dimensional self-attention-based speech enhancement

Requirements

Tensorflow = 1.4.0

Datasets

The speech stored in this git is enhanced by our 2D-SA. We evaluate the performance on two datasets.
(1) An open-source dataset [1].
(2) A large-scale dataset (Designed and generated by ourselves).

In this Git, enhanced speech, models, and the enhanced edges are uploaded.
The details and scripts of training and testing are included in folder Scripts
In the folder appendix, more explanations about model structure and hyper-parameters will be added.

Contact

If you have questions please contact: Email: lichenxing007@gmail.com

References:
[1] Cassia Valentini-Botinhao, Xin Wang, Shinji Takaki, and Junichi Yamagishi, “Investigating rnn-based speech enhancement methods for noise-robust text-to-speech,” in 9th ISCA Speech Synthesis Workshop, pp. 146–152.

More

A Pytorch implemention will be released soon.