Pushing the envelope - aside [speech recognition

Abstract

Despite successes, there are still significant limitations to speech recognition performance, particularly for conversational speech and/or for speech with significant acoustic degradations from noise or reverberation. For this reason, authors have proposed methods that incorporate different (and larger) analysis windows, which are described in this article. Note in passing that we and many others have already taken advantage of processing techniques that incorporate information over long time ranges, for instance for normalization (by cepstral mean subtraction as stated in B. Atal (1974) or relative spectral analysis (RASTA) based in H. Hermansky and N. Morgan (1994)). They also have proposed features that are based on speech sound class posterior probabilities, which have good properties for both classification and stream combination.

Keywords

Affiliated Institutions

• University of California, Berkeley US
• International Computer Science Institute US
• Columbia University US
• SRI International US
• Texas Instruments (United States) US
• Idiap Research Institute CH
• National Institute of Standards and Technology US
• University of Washington US
• Nokia (Finland) FI
• École Polytechnique Fédérale de Lausanne CH
• National Institute of Standards EG

Related Publications