The use of Maximum Entropy principle in continuous speech recognition

Research Student: Yasser H Abdel-Haleem Supervisors: Neil Lawrence and Steve Renals

A research project funded by Motorola.

The aim of this project is to develop acoustic modelling component based on the Maximum entropy (MaxEnt) principle. A continuous feature formulation of the maximum entropy modelling is utilized to model the acoustic variability. The suggested approach has two steps:

  1. the selection of the best acoustic space that efficiently and economically represent the acoustic data and its variability;
  2. the combination of selected acoustic features in the maximum entropy framework to estimate the posterior probabilities over the phonetic labels given the acoustic input.