COST278 and ISCA Tutorial and Research Workshop (ITRW) on Robustness Issues in Conversational Interaction

University of East Anglia, Norwich, UK
August 30-31, 2004

Environmental Robustness in Automatic Speech Recognition

Richard Rose

Dept. of Electrical and Computer Engineering, McGill University, Montreal, QC, Canada

The next generation of telecommunications networks promises to provide users with an array of services for providing automated access to sources of online information to be offered over personalized networks. The added mobility provided by this infrastructure will place users in an infinite variety of noisy acoustic environments. While there are many sources of variability that can impact the performance of automatic speech recognition (ASR) systems, the huge variety of potential background events and distortions that can occur in acoustic environments can have a particularly severe impact on ASR performance. This paper describes work that has been performed to address this problem. It also attempts to identify shortcomings in existing approaches and identify where breakthroughs are necessary to advance the current state of the art.

Full Paper

Bibliographic reference.  Rose, Richard (2004): "Environmental robustness in automatic speech recognition", In Robust2004, paper KRR.