Représentation et Estimation de la Force de Voix à partir du Spectre Moyen à Long Terme

Abstract : Representing and Recovering Voice Strength from the Long Term Average Spectrum The goal of the study is to recover the Sound Pressure Level emitted by a speaker, from the single long term spectrum envelope. The data consists of a set of 1/3rd octave Long Term Average Spectra, calibrated in sound level and exhibiting a large variability according to the speaker's gender, age and requested vocal effort degree. The visual representation of the spectra shows that it is more coherent to group them according to the emitted sound level than from the requested vocal effort degree. A comparison procedure is then applied to the data, after normalization of the spectra to a common, arbitrary value of their sound level. The results indicate that the single spectral envelope is sufficient to recover the emitted sound level, within a statistical margin of error smaller than 5 dB.
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01871854
Contributor : Jean-Sylvain Lienard <>
Submitted on : Tuesday, September 11, 2018 - 1:21:33 PM
Last modification on : Saturday, March 16, 2019 - 1:55:43 AM
Document(s) archivé(s) le : Wednesday, December 12, 2018 - 2:51:29 PM

File

JEP18.pdf
Publisher files allowed on an open archive

Identifiers

Citation

Jean-Sylvain Liénard. Représentation et Estimation de la Force de Voix à partir du Spectre Moyen à Long Terme. XXXe Journées d'Etudes sur la Parole, International Speech Communication Association, Jun 2018, Aix en Provence, France. ⟨10.21437/jep.2018-71⟩. ⟨hal-01871854⟩

Share

Metrics

Record views

32

Files downloads

39