This paper presents a technique for modelling the aspiration noise produced during phonation. The method employs the widely used LF voice source model, which is here extended to include a turbulence noise source for the generation of aspiration. Drawing on speech production theory and on empirical data, the overall amplitude level as well as the within-pulse modulation of the noise are determined by the specific shape of the LF model waveform. The main advantage of this approach is that it simplifies the control of the glottal source, as the temporal dynamics of the variation in noise level need not be explicitly set, but is generated automatically. Informal listening tests also suggest that the proposed modelling technique may contribute to the naturalness of synthesised speech, particularly with regard to breathy voicing.
Cite as: Gobl, C. (2006) Modelling aspiration noise during phonation using the LF voice source model. Proc. Interspeech 2006, paper 1718-Tue1FoP.4, doi: 10.21437/Interspeech.2006-310
@inproceedings{gobl06_interspeech,
author={Christer Gobl},
title={{Modelling aspiration noise during phonation using the LF voice source model}},
year=2006,
booktitle={Proc. Interspeech 2006},
pages={paper 1718-Tue1FoP.4},
doi={10.21437/Interspeech.2006-310}
}