Edinburgh Research Explorer

Speech Recognition using Linear Dynamic Models

Research output: Contribution to journalArticle

Related Edinburgh Organisations

Original languageEnglish
Pages (from-to)246-256
Number of pages11
JournalIEEE Transactions on Audio, Speech and Language Processing
Volume15
Issue number1
DOIs
Publication statusPublished - 1 Jan 2007

Abstract

The majority of automatic speech recognition (ASR) systems rely on hidden Markov models, in which Gaussian mixtures model the output distributions associated with sub-phone states. This approach, whilst successful, models consecutive feature vectors (augmented to include derivative information) as statistically independent. Furthermore, spatial correlations present in speech parameters are frequently ignored through the use of diagonal covariance matrices. This paper continues the work of Digalakis and others who proposed instead a first-order linear state-space model which has the capacity to model underlying dynamics, and furthermore give a model of spatial correlations. This paper examines the assumptions made in applying such a model and shows that the addition of a hidden dynamic state leads to increases in accuracy over otherwise equivalent static models. We also propose a time-asynchronous decoding strategy suited to recognition with segment models. We describe implementation of decoding for linear dynamic models and present TIMIT phone recognition results.

ID: 2077460