The aim of this study is to assess the potential for deliberately changing
one’s voice as a means to conceal or falsify identity, comparing
acoustic and perceptual measurements of carefully controlled speech
productions.
Twenty-two non expert speakers read a phonetically-balanced text
5 times in various conditions including natural speech, free vocal
disguise (2 disguises per speaker), impersonation of a common target
for all speakers, impersonation of one specific target per speaker.
Long-term average spectra (LTAS) were computed for each reading and
multiple pairwise comparisons were performed using the SDDD dissimilarity
index.
The acoustic analysis showed that all speakers were able to deliberately
change their voice beyond self-typical natural variation, whether in
attempting to simply disguise their identity or to impersonate a specific
target. Although the magnitude of the acoustic changes was comparable
in disguise vs. impersonation, overall it was limited in that it did
not achieved between-speaker variation levels. Perceptual judgements
performed on the same material revealed that naive listeners were better
at discriminating between impersonators and targets than at simply
detecting voice disguise.
Cite as: Delvaux, V., Caucheteux, L., Huet, K., Piccaluga, M., Harmegnies, B. (2017) Voice Disguise vs. Impersonation: Acoustic and Perceptual Measurements of Vocal Flexibility in Non Experts. Proc. Interspeech 2017, 3777-3781, doi: 10.21437/Interspeech.2017-1080
@inproceedings{delvaux17_interspeech,
author={Véronique Delvaux and Lise Caucheteux and Kathy Huet and Myriam Piccaluga and Bernard Harmegnies},
title={{Voice Disguise vs. Impersonation: Acoustic and Perceptual Measurements of Vocal Flexibility in Non Experts}},
year=2017,
booktitle={Proc. Interspeech 2017},
pages={3777--3781},
doi={10.21437/Interspeech.2017-1080}
}