Voice Source Contribution to Prominence Perception: Rd Implementation

Andy Murphy, Irena Yanushevskaya, Ailbhe Ní Chasaide, Christer Gobl

This paper explores the contribution of voice source modulation to the perception of prominence, following on previous analyses of accentuation, focus and deaccentuation. A listening test was carried out on a sentence of Irish with three accented, prominent syllables (P1, P2, P3). Using inverse filtering and resynthesis, a ‘flattened’ version was generated, with only slight declination of f0 and other voice source parameters. The global waveshape parameter Rd was modulated to provide (i) source boosting (tenser phonation) on either P1 or P2 and/or (ii) source attenuation (laxer phonation) following (Post-attenuation) or preceding (Pre-attenuation) P1 or P2. Rd variation was achieved in two different ways to generate two series of stimuli. f0 was not varied in either series. Twenty-nine listeners rated the prominence level of all syllables in the utterance. Results show that the phrasal position (P1 vs. P2) makes a large difference to prominence judgements. P1 emerged as overall more prominent and more readily ‘enhanced’ by the source modifications. Post-attenuation was particularly important for P1, with effects equal to or greater than local P-boosting. In the case of P2, Pre-attenuation was much more important than Post-attenuation.

 DOI: 10.21437/Interspeech.2018-2352

Cite as: Murphy, A., Yanushevskaya, I., Ní Chasaide, A., Gobl, C. (2018) Voice Source Contribution to Prominence Perception: Rd Implementation. Proc. Interspeech 2018, 217-221, DOI: 10.21437/Interspeech.2018-2352.

  author={Andy Murphy and Irena Yanushevskaya and Ailbhe {Ní Chasaide} and Christer Gobl},
  title={Voice Source Contribution to Prominence Perception: Rd Implementation},
  booktitle={Proc. Interspeech 2018},