Diffusion Posterior Sampling for Informed Single-Channel Dereverberation
This website contains supplementary material to the paper:
Code
Code for the proposed DPS and StateDPS is soon availabe at https://github.com/sp-uhh/derevdps
Code for the baseline algorithm SGMSE+ (originally proposed in Speech Enhancement and Dereverberation with Diffusion-Based Generative Models) is availabe at https://github.com/sp-uhh/sgmse
Code for the baseline algorithm StoRM (originally proposed in StoRM: A Diffusion-based Stochastic Regeneration Model for Speech Enhancement and Dereverberation) is availabe at https://github.com/sp-uhh/storm
No measurement noise
440o020f.wav | 442c0202.wav | 443c0207.wav | 445c020g.wav | |
---|---|---|---|---|
T60 | 0.43 s | 1.06 s | 1.96 s | 0.79 s |
Clean | ||||
Noisy | ||||
DPS [1] | ||||
StateDPS [1] | ||||
RIF+Post [2] | ||||
SGMSE+M [3, 4] | ||||
StoRM [5] |
Masurement noise: SNR = 10dB
440o020f.wav | 442c0202.wav | 440o020f.wav | 442c0202.wav | 440o020f.wav | 442c0202.wav | |
---|---|---|---|---|---|---|
T60 | 0.43 s | 1.06 s | 0.43 s | 1.06 s | 0.43 s | 1.06 s |
Noise type | Gaussian | Gaussian | CHiME3 [4] | CHiME3 [4] | Wind [7] | Wind [7] |
Clean | ||||||
Noisy | ||||||
DPS [1] | ||||||
DPS+Post [1, 8] | ||||||
RIF+Post [2] |
References
[1] J-M. Lemercier, S. Welker and T. Gerkmann. Diffusion Posterior Sampling for Single-Channel Informed Dereverberation. arXiv preprint arXiv:2306.12286. 2023.
[2] I. Kodrasi, T. Gerkmann, and S. Doclo, Frequency-domain single-channel inverse filtering for speech dereverberation: Theory and practice. ICASSP. 2014.
[3] J. Richter, S. Welker, J-M. Lemercier, B. Lay, and T. Gerkmann. Speech Enhancement and Dereverberation with Diffusion-Based Generative Models. IEEE TASLP. 2023.
[4] J-M. Lemercier, J. Richter, S. Welker and T. Gerkmann. Analysing Diffusion-based Generative Approaches versus Discriminative Approaches for Speech Restoration. ICASSP. 2023.
[5] J-M. Lemercier, J. Richter, S. Welker and T. Gerkmann. StoRM: A Diffusion-based Stochastic Regeneration Model for Speech Enhancement and Dereverberation. arXiv preprint arXiv:2212.11851. 2022.
[6] J. Barker, R. Marxer, E. Vincent and S. Watanabe. The third ‘CHiME’ speech separation and recognition challenge: Analysis and outcomes. Computer Speech & Language. 2016.
[7] J-M. Lemercier, J. Thiemann, R. Koning and T. Gerkmann. Wind Noise Reduction with a Diffusion-based Stochastic Regeneration Model. arXiv preprint arXiv:2306.12867. 2023
[8] C. Breithaupt, T. Gerkmann, and R. Martin. A novel a priori SNR estimation approach based on selective cepstro-temporal smoothing. ICASSP. 2008.