Speech Recognition with Dynamic Bayesian Networks

DynamicBayesian networks (DBNs) are a useful tool for representing complexstochastic processes. Recent developments in inference and learning in DBNsallow their use in real-world applications. In this paper, we apply DBNs to the problem of speech recognition. The factored state representation enabled by DBNsallows us to explicitly represent long-term articulatory and acoustic context in addition to the phonetic-state information maintained by hidden Markov models (HMMs). Furthermore, it enables us to modelthe short-term correlations amongmultiple observation streams within single time-frames. Given a DBN structure capable of representing these long- and short-term correlations, we applied the EMalgorithm to learn models with up to 500, 000 parameters. The use of structured DBN models decreased the error rate by 12 to 29% on a large-vocabulary isolated-word recognition task, compared to a discrete HMM; it also improvedsignificantly on other published results for the same task. This is the first successful application of DBNs to a largescale speech recognition problem. Investigation of the learned modelsindicates that the hidden state variables are strongly correlated with acoustic properties of the speech signal.

Authors

• Geoffrey Zweig Stuart Russell

Keywords

No keywords are indexed for this paper.