Improved HMM Models for High Performance Speech Recognition

Abstract

In this paper we report on the various techniques that we tmplemented in order to improve the basic speech recognition performance of the BYBLOS system. Some ot these methods are new, while others are not. We present methods that improved pertbrmance as well as those that did not. The methods include Linear Discrirninant Analysis, Supervised Vector Quantization, Shared Mixture VQ. Deleted Estimation of Context Weights, MMI Estimation Using "N-Best" Alternatives, CrossWord Triphone Models. While we have not yet combined all of the methods in one system, the overall word recognition error rate on the May 1988 test set using the Word-Pair grammar has decreased from 3.4% to 1.7%.