Third International Conference on Spoken Language Processing (ICSLP 94) Yokohama, Japan September 18-22, 1994

Statistical Trajectory Models for Phonetic Recognition

William D. Goldenthal, James R. Glass

Spoken Language Systems Group, Laboratory for Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA

In this work, we apply statistical trajectory models (STM's) to the task of phonetic recognition. STM's attempt to capture the dynamic characteristics and statistical dependencies of acoustic attributes in a segment-based framework. The approach is based on the creation of a track, fa, for each phonetic unit a. The track serves as a model of the dynamic trajectories of the acoustic attributes over the segment. The statistical framework for scoring incorporates the auto- and cross-correlation properties of the track error over time, within a segment. This paper presents the results of a series of phonetic recognition experiments using the timit acoustic-phonetic corpus [1]. Using the NIST train and core test sets we obtained context-independent and context-dependent recognition accuracies of 64.0% and 69.0% respectively.

Full Paper

Bibliographic reference.  Goldenthal, William D. / Glass, James R. (1994): "Statistical trajectory models for phonetic recognition", In ICSLP-1994, 1871-1874.