Up to CURRENT RESEARCH PROJECTS

Improving the Phonetic Discrimination of Hidden Markov Model Based Speech Recognisers Project

Steve Young, Phil Woodland, Valtcho Valtchev

The aim of this project is to investigate the use of an integrated approach to discriminative training of HMMs to improve their basic acoustic discrimination capability. The project is focusing on the use of local discriminative input transformations; a variety of discriminative objective functions trained by a back-propagation algorithms and the use of sub-phone models. Sponsor: EPSRC


mavis (hyping by svr-www-admin)@eng.cam.ac.uk
Tue Jul 19 23:35:21 BST 1994