Second International Conference on Spoken Language Processing (ICSLP'92)
Banff, Alberta, Canada
In distinction to phonated speech, in whispered speech the vocal cords do not vibrate and the quasi-periodic vibration with harmonics present in phonation is replaced by noise with continuous spectrum. However, it is not just the presence or absence of vocal cord vibration which distinguishes voicing from non-voicing. There are secondary voicing cues which make perceptual discrimination between whispered voiced and voiceless speech sounds possible (e.g., Chen and Nabelek, 1986). However, the number of discrimination errors in the whispered speech is higher than in the phonated speech. In whisper, the largest number of errors occurred in cognates. In those, the place of articulation was, in general, identified correctly but the voicing feature was often misidentified. Still, the distinction between voicing and non-voicing was significant. Spectrographic analysis showed differences in formant trajectories, onset rates of bursts following the release, energy distributions, and others, that might serve as cues in voicing identification.
Bibliographic reference. Nabelek, Igor V. (1992): "Coding of voicing in whispered plosives", In ICSLP-1992, 1119-1122.