Estimating acoustic speech features in low signal-to-noise ratios using a statistical framework

Abstract

Accurate estimation of acoustic speech features from noisy speech and from different speakers is an ongoing problem in speech processing.