Evaluating speech samples designed for the Voice Profile Analysis Scheme for Brazilian Portuguese (BP-VPAS)

Abstract

The present study aims at evaluating a corpus designed for voice quality settings analysis (Laver, 1980). It addresses the role of keyspeech segments for the identification of voice quality settings. The corpus, recorded by 60 speakers, contained repetitions of 10 keyspeech sentences, and semi-spontaneous speech samples. These samples were evaluated by two expert subjects according to the BP-VPAS Profile (Camargo, Madureira, 2008). These data were analyzed by means of an artificial neural network and statistically by multiple regression analysis. The data reinforce the importance of the description of voice quality based on the phonetic model and the relevant role of keyspeech segments.