Skip to Main content Skip to Navigation

Etude de la source glottique en voix parlée et chantée : modélisation et estimation, mesures acoustiques et électroglottographiques, perception

Abstract : The dependence of qualities of the voice on parameters of glottal flow was studied in speech and singing, using spectral modelling and estimation, acoustic and electroglottographic measurements and psychoacoustic determinations of just noticeable differences (JND). The effects of glottal flow parameters on the voice signal were studied in the time and frequency domains using models of glottal flow. An increase of the amplitude parameters led to a spectral amplification. The open quotient and asymmetry coefficient were correlated to the frequency and bandwith of the glottal formant and the spectral tilt amplitude. The return phase quotient was correlated to the high-frequency part of the spectral tilt. Open quotient estimation using the first two harmonics difference, H1-H2, was studied theoretically and experimentally. This estimation is not adequate for singing, for which the interaction between source and filter is no longer negligible. The use of the derivative of electroglottographic signals to detect glottal opening and closure instants, independently of the effects of the supraglottal system, was investigated and methods were proposed for measuring fundamental frequency and open quotient. A data base of simultaneous EGG and acoustic signals was recorded for 18 trained singers who produced sustained vowels, crescendos and spoken and sung phrases. The open quotient was analysed as a function of fundamental frequency and vocal intensity and related to the main laryngeal mechanisms of voice production. It had larger values in mechanism II (0.5 < Oq < 0.95) than in mechanism I (0.3 < Oq < 0.8) and a jump of open quotient was noticed at the mechanism's transition. A high correlation was found with vocal intensity in mechanism I and with fundamental frequency in mechanism II. It was also dependent on the opening of the vowel. The JNDs were determined for the open quotient and the asymmetry coefficient. The relative JND, DOq/Oq, was about 14 % for 20 untrained subjects and 10 % for 10 trained ones. The JND did not vary with a change of vowel or fundamental frequency, but it depended on the amplitude parameter and whether or not a vibrato were present.
Complete list of metadatas
Contributor : Nathalie Henrich Bernardoni <>
Submitted on : Monday, January 8, 2007 - 2:35:32 PM
Last modification on : Thursday, December 10, 2020 - 12:32:45 PM
Long-term archiving on: : Tuesday, April 6, 2010 - 8:02:18 PM


  • HAL Id : tel-00123133, version 1


Nathalie Henrich Bernardoni. Etude de la source glottique en voix parlée et chantée : modélisation et estimation, mesures acoustiques et électroglottographiques, perception. Acoustique [physics.class-ph]. Université Pierre et Marie Curie - Paris VI, 2001. Français. ⟨tel-00123133⟩



Record views


Files downloads