Overview of Current Text-to-Speech Techniques: Part II - Prosody and Speech Generation
MD EDGINGTON, A. Lowry, Philip J. B. Jackson, Andrew Breen, Stephen Minnis · UEA Digital Repository (University of East Anglia) · 1996
Speech synthesis is an increasingly important technique in interactive voice systems, and is crucial to the next generation of spoken language dialogue systems. Described here are many of the techniques and algorithms used in modern commercial text-to-speech synthesisers, where real world issues of robustness, reliability and processing cost compete with the specific requirements of naturalness and intelligibility. This paper concentrates on using the results of the linguistic analysis of the input text to generate an appropriate synthetic prosody and speech signal. The importance of the rhythmic characteristics of synthetic speech is emphasised by placing the prosody generation task within the framework of metrical phonology.