This paper investigates the use of Gaussian Selection (GS) to increase the speed of a large vocabulary speech recognition system. Typically 30-70% of the computational time of a continuous density HMM-based speech recogniser is spent calculating probabilities. The aim of GS is to reduce this load by selecting the subset of Gaussian component likelihoods that should be computed given a particular input vector. This paper examines new techniques for obtaining "good" Gaussian subsets or "shortlists". All the new schemes make use of state information, specifically which state each of the Gaussian components belongs to. In this way a maximum number of Gaussian components per state may be specified, hence reducing the size of the shortlist. The first technique introduced is a simple extension of the standard GS method, which uses this state information. Then, more complex schemes based on maximising the likelihood of the training data are proposed. These new approaches are compared with the standard GS scheme on a large vocabulary speech recognition task. On this task, the use of state information reduced the percentage of Gaussians computed to 10-15%, compared with 20-30% for the standard GS scheme, with little degradation in performance. 1 M.J.F.Gales is now at the IBM T.J. Watson Research Center, Yorktown Heights, NY 10598, USA. 2 K.M. Knill is now at Nuance Communications, 1380 Willow Rd, Menlo Park, CA 94025, USA. List of Tables 1 Change in the average forced alignment likelihood of the ARPA 1994 H1 development data for SGS and SBGS systems, compared to the standard no GS system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2 Recognition performance of the standard no GS, SGS and SBGS systems on the ARPA 1994 H...

In this paper we report on some recent improvements on the acoustic modeling in a segment-based speech recognition system. Context-dependent segment models and improved pronunciation modeling are shown to reduce word error rates in a telephone -based, conversational system by over 18%, while the technique of Gaussian selection reduces overall computation by more than a factor of two.

This thesis explores the use of discriminative training to improve acoustic modeling in a segment-based speech recognizer. In contrast with the more commonly used Maximum Likelihood training, discriminative training considers the likelihoods of competing classes when determining the parameters for a given class's model. Thus, discriminative training works directly to minimize the number of errors made in the recognition of the training data.

The geometric distribution of states' duration is one of the main performance limiting assumptions of hidden Markov modeling of speech signals. Stochastic segment models, generally, and segmental HMM, specifically, overcome this deficiency partly at the cost of more complexity in both training and recognition phases. In addition to this assumption, the gradual temporal changes of speech statistics has not been modeled in HMM. In this paper, a new duration modeling approach is presented. The main idea of the model is to consider the effect of adjacent segments on the probability density function estimation and evaluation of each acoustic segment. This idea not only makes the model robust against