Dr. Stefanie Hufnagel

The Speech Communication Group at RLE studies speech processing in human listeners and aims to model those processes in automatic speech analysis systems. Three key projects include 1) LEXI (Linguistic Events eXtraction and Integration) – a speech recognition system based on abrupt acoustic events called Landmarks, 2) Automatic analysis of speech produced by both typically-developing and atypically-developing children, and 3) Comparison of patterns of noisy speech recognition by humans and by ASR (Automatic Speech Recognition) systems. The goal in all three projects is to incorporate insights gained from our growing understanding of the basic science of human speech processing into automatic systems designed for various uses. Our research has applications in the design of improvements for current ASR systems (which are often challenged by noisy speaking environments, new words and new speakers) and in standardized clinical analysis for automatic diagnosis of speech disabilities, as well as in second language learning. Preliminary evidence suggests that this landmark-based approach is applicable not just for English, but across human languages."