Computational aeroacoustics strategies for the simulation of voice sequences involving sibilants - Forum Acusticum 2020 Access content directly
Conference Papers Year : 2020

Computational aeroacoustics strategies for the simulation of voice sequences involving sibilants

Abstract

Realistic finite element (FEM) simulations of voice se-quences containing sibilant consonants like /sa/, /za/ or/asa/, to mention a few, turns to be a very demanding prob-lem from a computational point of view. The sole re-production of a single sibilant implies resorting to com-putational aeroacoustics (CAA) strategies. One can ei-ther employ a direct numerical simulation (DNS) of thefull compressible Navier-Stokes equations or use a hy-brid approach in which the incompressible or compress-ible Navier-Stokes equations are first solved to obtain anacoustic source term from the flow jet motion. After that,an acoustic wave operator is used to synthesize the sibi-lant sound. If the sibilant is part of a voice sequence thatincludes vowels, the whole problem needs to be solved ina moving vocal tract (VT) geometry and a train of glottalpulses must be prescribed at the glottis for vowel genera-tion. Dealing with dynamic VTs leads the computationalcost of the problem prohibitive, even if one resorts to su-percomputer facilities.In this work, we review CAA approaches and some re-cently proposed alternatives to produce sibilants and voicesequences with sibilants at a reasonable computationalcost. The key idea is to approximate the acoustic flowsource term to avoid solving the Navier-Stokes equationsin the first step of CAA. Two options are considered. Thefirst one consists in using a random distribution of spinningKirchhoff’s vortices to emulate the quadrupole acousticsources between the lower incisors and lower lips. The sec-ond one simply uses a white noise amplitude monopole tosimulate the interdental flow jet and a dipole to account forincisors’ diffraction effects. The wave equation in mixedform is solved to produce the final sounds. Numerical andimplementation details are outlined, as well as the pros andcons of the proposed options. Some videos and audios ofnumerically generated voice sequences will be shown inthe conference presentation.
Fichier principal
Vignette du fichier
000145.pdf (324.25 Ko) Télécharger le fichier
Origin : Explicit agreement for this submission

Dates and versions

hal-03221386 , version 1 (21-05-2021)

Identifiers

Cite

Oriol Guasch, Marc Arnela, Arnau Pont. Computational aeroacoustics strategies for the simulation of voice sequences involving sibilants. e-Forum Acusticum 2020, Dec 2020, Lyon, France. pp.1345-1348, ⟨10.48465/fa.2020.0145⟩. ⟨hal-03221386⟩

Collections

FA2020
54 View
57 Download

Altmetric

Share

Gmail Facebook X LinkedIn More