Correcting the reverse process
This website contains supplementary material to the paper:
- Single and Few-step Diffusion for Generative Speech Enhancement
Code here
| Filename | Clean | Noisy | BBED 5 NFEs [1] | Predictive baseline | proposed CRP 1 NFE | proposed CRP 5 NFEs |
|---|---|---|---|---|---|---|
| 447c0201.wav | ||||||
| 445o030a.wav | ||||||
| 445c0202.wav | ||||||
| 445o030y.wav | ||||||
| 444c020g.wav |
| Filename | Clean | Noisy | BBED 5 NFEs [1] | Predictive baseline | proposed CRP 1 NFE | proposed CRP 5 NFEs |
|---|---|---|---|---|---|---|
| 447c0201.wav | ||||||
| 445o030a.wav | ||||||
| 445c0202.wav | ||||||
| 445o030y.wav | ||||||
| 444c020g.wav |
References
[1] Bunlong Lay, Simon Welker, Julius Richter, Timo Gerkmann "Reducing the Prior Mismatch of Stochastic Differential Equations for Diffusion-based Speech Enhancement", ISCA Interspeech, Dublin, Ireland, Aug. 2023