See discussions, stats, and author profiles for this publication at:Waveguide physical modeling of vocal tract acoustics: Flexible formantbandwidth control from increased model dimensionalityArticleinIEEE Transactions on Audio Speech and Language Processing · June 2006DOI: 10.1109/TSA.2005.858052 · Source: IEEE XploreCITATIONS48READS2433 authors, including:Some of the authors of this publication are also working on these related projects:Singing behaviour and development across the lifespanView projectWidening Young Male Participation in ChorusView projectDavid M HowardRoyal Holloway, University of London393PUBLICATIONS2,922CITATIONSSEE PROFILEAll content following this page was uploaded byDavid M Howardon 23 January 2015.The user has requested enhancement of the downloaded file.
promoting access to White Rose research papersWhite Rose Research OnlineUniversities of Leeds, Sheffield and YorkWhite Rose Research Online URL for this paper:Published paperMullen, J., Howard, D.M. and Murphy, D.T. (2006)Waveguide Physical Modelingof Vocal Tract Acoustics: Flexible Formant Bandwidth Control From IncreasedModel Dimensionality,IEEE Transactions on Audio, Speech and LanguageProcessing, Volume 14 (3), 964 - 971.[email protected]
964IEEE TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING, VOL. 14, NO. 3, MAY 2006Waveguide Physical Modeling of Vocal TractAcoustics: Flexible Formant Bandwidth ControlFrom Increased Model DimensionalityJack Mullen, David M. Howard, and Damian T. MurphyAbstract—Digital waveguide physical modeling is often usedas an efficient representation of acoustical resonators such as thehuman vocal tract. Building on the basic one-dimensional (1-D)Kelly–Lochbaum tract model, various speech synthesis techniquesdemonstrate improvements to the wave scattering mechanismsin order to better approximate wave propagation in the complexvocal system. Some of these techniques are discussed in this paper,with particular reference to an alternative approach in the form ofa two-dimensional waveguide mesh model. Emphasis is placed onits ability to produce vowel spectra similar to that which would bepresent in natural speech, and how it improves upon the 1-D model.Tract area function is accommodated as model width, rather thantranslated into acoustic impedance, and as such offers extra controlas an additional bounding limit to the model. Results show thatthe two-dimensional (2-D) model introduces approximately linearcontrol over formant bandwidths leading to attainable realisticvalues across a range of vowels. Similarly, the 2-D model allows forapplication of theoretical reflection values within the tract, whichwhen applied to the 1-D model result in small formant bandwidths,and, hence, unnatural sounding synthesized vowels.