Reproduce “Impulse Noise Removal In Speech Using Wavelets” 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
R. C. Nongpiur
A method for removing impulse noise from speech signals leverages wavelet transforms to differentiate between speech components and noise based on their distinct decay rates and regularity. This approach effectively suppresses impulse noise from car speech recordings corrupted by rain sounds, preserving speech quality and avoiding audible artifacts.
Key insights
- The multi-resolution nature of wavelet transforms enables simultaneous high time resolution for transient noise and high frequency resolution for speech, overcoming limitations of prior methods.
- Impulse noise, vowels, and consonants exhibit distinct Lipschitz regularity and wavelet coefficient decay patterns across scales, allowing for their differentiation.
- Dynamic thresholding, derived from a median-filtered local speech envelope in the wavelet domain, effectively targets impulses without impacting underlying speech.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.