ModProso

ModProso is a Praat-based tool for prediction and modification of F0 contours in natural utterances.

It was developed as a research tool to evaluate the perceptual validity of the automatic F0 stylisation, annotation and modelling representations obtained with MelAn, but it can also be used to generate synthetic stimuli with modified F0 contours for several perception experiments.

ModProso is a language-dependent tool: it was originally designed for its use with speech utterances in Spanish and Catalan, but current research in being carried out to adapt it to Brazilian Portuguese. Adaptation to other languages could be also done with a minimum effort.

ModProso performs basically two tasks:

• • the prediction of a chain of F0 target values, in the form of a Praat-style stylised contour, using as input a set of local pattern labels and two P and V reference lines.

  • the substitution of the original F0 contour by the predicted one in the speech utterance provided as input.

To perform these two tasks, ModProso needs as input:

• • a wav file containing the utterance to be manipulated

  • a Textgrid file containing the orthographic and phonetic transcription of the provided utterance, and its prosodic segmentation into syllables, SG, IG and breath groups (BG)

  • a 'contour' text file containing the melodic labels

  • a couple of 'regression_P' and 'regression_V' files containing the values for the global reference F0 lines

At the end of the process, an edition window showing the speech signal and the obtained stylised contour is displayed. Using this window, the user can obtain a synthesised version of input utterance using Overlap-Add or LPC techniques, which can be played directly or stored in an output wav file.