Discriminative mixture weight estimation for large Gaussian mixture models

This paper describes a new approach to acoustic mod-eling for large vocabulary continuous speech recognition (LVCSR) systems. Each phone is modeled with a large Gaussian mixture model (GMM) whose context-dependent mixture weights are estimated with a sentence-level discrim-inative training criterion. The estimation problem is casted in a neural network framework, which enables the incorporation of the appropriate constraints on the mixture weight vectors, and allows a straightforward training procedure, based on steepest descent. Experiments conducted on the Callhome-English and Switchboard databases show a signiicant improvement of the acoustic model performance, and a somewhat lesser improvement with the combined acoustic and language models .