Abstract

An unsupervised, discriminative, sentence level, HMM adaptation based on speech-silence classi? cation is pre sented. Silence and speech regions are determined either using a speech end-pointer or the segmentation obtained from the recognizer in a? rst pass. The discriminative training procedure using a GPD or any other discriminative training algorithm, employed in conjunction with the HMM based recognizer, is then used to increase the discrimination between silence and speech.

Date

June 13, 2000

Authors

SS Narayanan, A Potamianos, I Zeljkovic

Inventors

Shrikanth Sambasivan Narayanan, Alexandros Potamianos, Ilija Zeljkovic

Patent_office

US

Patent_number

6076057

Application_number

08861413