The validity of some acoustic measures to predict voice quality settings: trends between acoustic and perceptual correlates of voice quality

Abstract

The present study aims at investigating the power of acoustic measures to predict voice quality settings detected perceptually. The corpus was composed by speech samples recorded by 60 subjects. The audio samples were analyzed in PRAAT by means of the SG Expression Evaluator Script (Barbosa, 2009), which extracts f0 and f0 first derivate, intensity, spectral tilt and long-term average spectrum measures. The same samples were perceptually evaluated by means of the Voice Profile Analysis Scheme for Brazilian Portuguese (BP-VPAS). The methodological procedures made it possible to identify the clusters related to the acoustic measures group and that of the perceptual judgement results (cluster analysis) as well as to correlate these two groups (canonic correlation analysis).