As a first step toward development of a "speech montage system", this paper attempts to derive a core set of Japanese epithets which are commonly used in an everyday life to represent voice quality features associated with talker individuality. Perceptual experiments were conducted, where subjects were asked to evaluate sentence utterances recorded from a variety of male speakers in terms of 25 epithets which were derived in another experiment [1] to be indicative of voice quality relevant to talker individuality. The evaluation scores were subjected to a statistical clustering analysis. The analysis resulted in that the 25 epithets could be grouped into either eight categories for male or seven for female subjects. These categories were basically the same as those obtained in the previous experiment [1] where subjects were required to evaluate their own voice with the same set of 25 epithets. Agreement between the results from the two experiments guarantees reliability of the core epithet categories to represent voice quality associated with talker individuality.
Cite as: Kido, H., Kasuya, H. (1998) Representation of voice quality features associated with talker individuality. Proc. 5th International Conference on Spoken Language Processing (ICSLP 1998), paper 1005, doi: 10.21437/ICSLP.1998-214
@inproceedings{kido98_icslp,
author={Hiroshi Kido and Hideki Kasuya},
title={{Representation of voice quality features associated with talker individuality}},
year=1998,
booktitle={Proc. 5th International Conference on Spoken Language Processing (ICSLP 1998)},
pages={paper 1005},
doi={10.21437/ICSLP.1998-214}
}