The ESCA Workshop on Speech Synthesis

September 25-28, 1990
Autrans, France

Deriving Text-To-Speech Durations from Natural Speech

Jan P. H. van Santen

Linguistics Research Department, AT&T Bell Laboratories, Murray Hill, NJ, USA

Any text-to-speech system has a subsystem (duration system) that computes speech timing. How does one construct a duration system that accurately mimics natural speech? This paper discusses a particular type of data analysis method for the statistical analysis of natural speech durations, ordinal data analysis, and shows how it can be used for the construction of duration systems.


Full Paper

Bibliographic reference.  Santen, Jan P. H. van (1990): "Deriving text-to-speech durations from natural speech", In SSW1-1990, 157-160.