Reproduce “Mel-McNet: A Mel-Scale Framework for Online Multichannel Speech Enhancement” 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
Yujie Yang, Bing Yang, Xiaofei Li
Mel-McNet reduces reported computation for multichannel speech enhancement by compressing microphone features into a perceptually motivated Mel-frequency representation while retaining spatial cues. On six-channel simulated CHiME-3, it uses 46.96 versus McNet’s 115.13 FLOPs per second, with both at 4.4% WER and DNSMOS scores of 3.14 versus 3.12. However, WB-PESQ and STOI are lower; results are limited to this benchmark.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.