AN OVERVIEW OF THE SPHINX SPEECH RECOGNITION SYSTEM

Speaker independence, continuous speech, and large vocabularies pose three of the greatest challenges in automatic speech recognition. Previously, accurate speech recognizers avoided dealing simultaneously with all three problems. This paper describes SPHINX, a system that demonstrates the feasibility of accurate, large-vocabulary speaker-independent, continuous speech recognition. SPHINX is based on discrete hidden Markov models (HMM's) with LPC-derived parameters. To provide speaker independence, we added knowledge to these HMM's in several ways: multiple codebooks of fixed-width parameters, and an enhanced recognizer with carefully designed models and word duration modeling. To deal with coarticulation in continuous speech, yet still adequately represent a large vocabulary, we introduce two new subword speech units—function-word-dependent phone models and generalized triphone models. With grammars of perplexity 997, 60, and 20, SPHINX attained word accuracies of 71, 94, and 96 percent on a 997-word task. © 1990 IEEE