Skip to content

Latest commit

 

History

History
15 lines (9 loc) · 486 Bytes

README.md

File metadata and controls

15 lines (9 loc) · 486 Bytes

DCTCRN

This is the unofficial implementation of DCTCRN, from paper Real-time Monaural Speech Enhancement With Short-time Discrete Cosine Transform.

Arxiv: https://arxiv.org/abs/2102.04629

Result

Model is trained on 28spk version of voicebank+demand dataset from here

Evaluation is has done on test set of voicebank+demand dataset.

MODEL PESQ STOI SI-SNR
DCTCRN-T 2.550 0.932 17.233