Estudo exploratório sobre a voz crepitante no português brasileiro

Abstract

This work seeks to understand the occurrence of creaky voice as a laryngeal adjustment in the linguistic context of speech, being an exploratory study that employs a qualitative approach to evaluate acoustic graphs, glottal signal by electroglottograph (EGG), and quantitative evaluations to classify the types of creakiness found and to understand the contexts that condition them. Laboratory Phonology is used as the analysis methodology based on quantitative methods of acoustic measurements in line with the qualitative analysis of language. By evaluating the simultaneous acoustic and glottal signal, it is possible to characterize in greater detail different types of voice and the occurrence of creaky voice. Production data from an experiment with 10 participants were evaluated, in which it was possible to find 514 occurrences of creaky voice. The qualitative analysis of the data obtained points to the existence of types of creaky voice distinct from the prototypical creaky voice, which may present different acoustic measurements than expected for prototypical creakiness. In the quantitative analysis, parameters of fundamental frequency (f0), spectral tilt (H∗ 1 − H∗ 2 , H∗ 2 − H∗ 4 e H2k − H5k), cepstral peak prominence (CPP), harmonic-to-noise ratio (HNR), and openness quotient (OQ) were evaluated. The results obtained indicate that creaky voice occurs in typical speech, with an average rate of 1.48%, with no predominance between men and women or prosodic position. There is a preference for creaky voice by the vowels [E], [a], and [u] and in tonic position, but this should be interpreted with caution. Samples with creaky voice exhibit lower f0, CPP,H∗ 1 − H∗ 2 , and H∗ 2 −H∗ 4 values than samples with modal voice, as expected. OQ values exhibit a bimodal distribution and, along with their variability, may indicate different types of creaky voice.