A New Intonation Model for Text-to-speech Synthesis

The text-to-speech intonation model we are developing derives from both linguistics, and the acoustics and aerodynamics of speech production. Our underlying premise is that in human speech production there are physical processes intrinsic to speech production, and that some of these processes can be cognitively represented – they can therefore become part of the domain of language processing. The model is based on our general philosophy of factoring out intrinsic and extrinsic physical phenomena to create associations between physical and cognitive representations. The model is easily extended to handle variability beyond the neutral rendering of intonation, using overlays to add pragmatically determined intentional and emotional effects.