Reproduce “Multi-Teacher Distillation for Cross-Domain Streaming Electrolaryngeal Speech Encoding” 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
Benedikt Mayrhofer, Enrique Orozco Olivares, Franz Pernkopf, Philipp Aichinger, Martin Hagmüller
The Mel-Conformer is a compact streaming encoder trained first on healthy speech and then with electrolaryngeal-aware recognition features, with optional dynamic time warping alignment. On German ELHE speech, its CTC-probe word error rate is 21.2% ± 3.0, versus 39.3% ± 2.8 for zero-shot WavLM large. This ASR-probe result does not establish voice-conversion quality or broad device readiness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.