Convolutional Recurrent MetriCGAN With Spectral Dimension Compression For Full-Band Speech Enhancement

Zhongshu Hou; Qinwen Hu; Tianchi Sun; Yuxiang Hu; Changbao Zhu; Kai Chen

ICASSP 2023accepted0 citations

Convolutional Recurrent MetriCGAN With Spectral Dimension Compression For Full-Band Speech Enhancement

Zhongshu Hou, Qinwen Hu, Tianchi Sun, Yuxiang Hu, Changbao Zhu, Kai Chen

Abstract

MetricGAN and its variations have been proven to be an effective wide-band speech enhancement model. In this paper, we expand it to full-band enhancement by combining our recently proposed learnable spectral dimension compression mapping strategy. The encoder-decoder structure with a time-frequency convolutional recurrent network is utilized as the generator. The proposed model is submitted to the ICASSP Signal Processing Grand Challenge: DNS-5 Challenge (2023). Without using the enrollment speech, it obtains a final score of 0.548 on Track-1 and 0.559 on Track-2.

BibTeX

@inproceedings{icassp2023_convolutionalrec,
  title = {Convolutional Recurrent MetriCGAN With Spectral Dimension Compression For Full-Band Speech Enhancement},
  author = {Zhongshu Hou and Qinwen Hu and Tianchi Sun and Yuxiang Hu and Changbao Zhu and Kai Chen},
  booktitle = {ICASSP 2023},
  year = {2023}
}