Fourth European Conference on Speech Communication and Technology

Madrid, Spain
September 18-21, 1995

A Dynamical System Model for Recognizing Intonation Patterns

Ken Ross, Mari Ostendorf

Boston University, Boston, MA, USA

This paper describes a new computational model of prosody for recognizing symbolic intonation patterns, specifically the different tones that mark pitch accents and phrase boundaries. The model represents intonation at multiple levels (segmental, syllable, and phrase levels) to capture acoustic feature dependence at different time scales. We take a probabilistic approach to intonation label recognition that utilizes a state-space dynamical system model at the syllable level. Recognition and training algorithms are described and results are reported from experiments on prosodic labeling of radio news speech.

Full Paper

Bibliographic reference.  Ross, Ken / Ostendorf, Mari (1995): "A dynamical system model for recognizing intonation patterns", In EUROSPEECH-1995, 993-996.