The ExProsodia research project started in 2006, based on the proposition of an intonation analysis model that especially considered the frequency variations above and below the accumulated average in the time of those frequencies. The values were then taken at three semitone intervals, which would characterize five significant tonal areas. This proposal, however, was not consistent with the attempts made to corroborate it. Although the mean tone obtained by the accumulated average over time of the frequencies proved to be quite stable to be taken as a reference, the fixed intervals above and below this average tone were not significant to be taken, also, as references.
From perception tests carried out within the project, the tonal areas could be simplified. There was motivation to widen the area corresponding to the mean tone to an interval of 7 semitones, regularly distributed: 3 semitones above the mean tone and 4 semitones below. Although the interval of 3 semitones above was always regular, the interval below had less regularity, oscillating between three and four semitones. The tonal areas that exceeded this range were broken down as Focus/Emphasis, above, and Finalization, below.
The software developed in Excel VBA that performed these calculations was then registered in 2010. This version made the analyzes based on the hypotheses of the mean tone, the Focus/Emphasis and the Finalization, as significant areas. The results obtained with these analyzes remained highly consistent with each other when applied to Portuguese-language data. When analyzing data from other languages, particularly English and Guarani, the results obtained did not maintain coherence. This difference required that new hypotheses be established, especially regarding the intervals characterizing tonal areas.
In 2018, a new version of the ExProsodia software was developed, in which it is possible to modify the defining intervals of tonal areas. Thus, the mean tone range could be treated from greater or lesser perceptual variations. As the medium tone is the reference for all other measurements made by the software, it was necessary that the whole set be reformulated in order to follow the needs of the speakers, without losing the initial hypotheses.
Based on a wide bibliography, the software was then reformulated in order to include two reference points: the global mean tone (the mean tone itself) and the local mean tone. The local mean tone takes as the starting point for its calculations the values established from the reset of the intonational phrases. This reset of the intonational phrases establishes specific segmentations in the frequency time series and restarts the process of calculating the accumulated average over time. The hypothesis of this local mean tone is that there is a correlation between each of these moments of reset of intonational phrases and the constitution of lexically considered phrases.
The possibility of obtaining two reference points - the global mean tone and the local mean tone — to interpret the intonation curve, associated with the proposed segmentations for reset of intonational phrases, allowed us to obtain a greater range of observable data capable of covering an equally greater spectrum of tonal variations conditioned by both intentional and unintentional expressive needs. In general, it is proposed that the variations in the global mean tone are unintentional, because they would be associated with the emotional conditions of the speaker and that the variations in the local mean tone are controlled by the speaker to obtain some type of expression desired. From this point of view, it is understood that at least these parameters can be used as grammatically predictable resources.