Reproduce “aTENNuate: Optimized Real-time Speech Enhancement with Deep SSMs on Raw Audio” 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
Yan Ru Pei, Ritik Shrivastava, FNU Sidharth
aTENNuate denoises raw-waveform speech using a state-space autoencoder, processing audio without a spectrogram front end. Its reported average wideband PESQ scores exceed DeepFilterNet3’s on VoiceBank + DEMAND and DNS1 synthetic no-reverberation tests, with fewer parameters but longer base latency. Removing PreConvs reduces listed latency from 46.5 to 16 ms, while lowering PESQ; these are theoretical network measures, not device benchmarks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.