F0 contour modelling

HUMBERTO MAXIMILIANO TORRES

Oro Verde, Entre Ríos

Workshop; Workshop on Advanced Signal Processing, Machine Learning and Modeling; 2022

Facultad de Ingeniería, Universidad Nacional de Entre Ríos

Intonation is one of the most important prosody attributes of natural speech, carrying linguistic, paralinguistic and non-linguistic information. Fundamental frequency (F0) contour and pauses are the two most important physical correlates of intonation. A meticulous description of the F0 contour is useful both for understanding its relationships to the underlying information and for its use in applications such as high-quality speech synthesis.I will introduce some of our works on modeling the F0 contour for Argentine Spanish. In particular, the works related to a model developed by Professor Hiroya Fujisaki, from my first contact with this model to the works that we are doing nowadays and those that we are planning for the near future. Fujisaki’s model of intonation has been successfully tested for different languages, it stands out for its simplicity and strong physiological basis. This model parameterizes F0 contours in an efficient manner: with a small number of parameters we can achieve a desired level of fitting accuracy. It is not my intention to discuss the philosophy behind this model, what I would like to raise are its physical and physiological foundations.