Features used to discriminate vowel height in voiced and whispered speech

Abstract

The purpose of this study was to define acoustic cues used to discriminate vowel height in voiced and whispered speech. Seventeen speakers produced sustained oral vowels, disyllabic words, sentences and read a phonetically balanced text. These tasks were repeated in voiced and whispered speech and analysed using the following parameters: Fundamental frequency, formant frequencies, spectral slope, sound pressure level and durations. Kernell density estimation plots and Pillai scores were used to characterise the vowel spaces and the degree of overlap between vowels. First formant frequency and relative duration were consistently used as height cues across the two speech modes (voiced and whispered). Whispered vowel spaces shifted downward (relative to voiced), and vowel pairs /i-a/, /a-ɔ/, /ɔ-u/ and /u-i/ were distinct when produced in both speech modes. The evidence presented can be used to restore voiced speech signals and to inform rehabilitation strategies.