Automatic detection of syllable nuclei as applied to segmentation of speech

This paper describes a speaker-independent algorithm which splits continuous speech utterances into intersyllabic segments. The intersyllabic segment is specified as the sound segment bounded by two successive syllable centers which typically correspond to the steady state portions of the syllable nuclei. Experiments with ten sentences spoken by two male and two female speakers showed fairly reasonable results. The intersyllabic segment is expected to be an effective unit for segmentation and labeling which takes advantage of some constraints on the sequencing of the sound segment features.