A Maximum Likelihood Approach to Continuous Speech Recognition

TLDR

Speech recognition is framed as a maximum‑likelihood decoding problem that requires statistical models of speech production. The paper aims to describe statistical models for speech recognition. The authors develop statistical models, focus on parameter estimation from sparse data, and present two decoding methods—one for constrained artificial languages and another for realistic tasks—illustrated by reviewing decoding results. The methods yield promising decoding results.

Abstract

Speech recognition is formulated as a problem of maximum likelihood decoding. This formulation requires statistical models of the speech production process. In this paper, we describe a number of statistical models for use in speech recognition. We give special attention to determining the parameters for such models from sparse data. We also describe two decoding methods, one appropriate for constrained artificial languages and one appropriate for more realistic decoding tasks. To illustrate the usefulness of the methods described, we review a number of decoding results that have been obtained with them.