Reproduce “VM-ARRAYDPS: Virtual Microphone Augmented Diffusion Posterior Sampling for Unsupervised Blind Speech Separation” 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
Jingqi Sun, Haozhan Tang, Shulin He, Zhong-Qiu Wang
VM-ArrayDPS guides diffusion sampling for blind speech separation using both physical microphone recordings and virtual observations derived from those same recordings by independent vector analysis. At each sampling step, it adds weighted reconstruction constraints for the virtual observations to ArrayDPS’s physical-microphone constraints. On simulated reverberant speech, its two-speaker test improves SI-SDR from 15.0 to 16.1 dB and SDR from 15.8 to 16.9 dB; this does not establish performance on real-room recordings or dereverberation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.