Reproduce “Diffusion Posterior Sampling for Informed Single-Channel Dereverberation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jean-Marie Lemercier, Simon Welker, Timo Gerkmann
The authors introduce a diffusion posterior sampling method for single-channel dereverberation that combines a learned clean-speech prior with a known room impulse response (RIR). During generation, the method uses a denoised speech estimate to guide consistency with the recording. In evaluation, it is more robust to added zero-mean Gaussian noise than informed inverse filtering, but assumes the RIR is known.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.