Glottal-wave and vocal-tract-area-function estimations from vowel sounds based on realistic assumptions and models
Huiqun Deng, Rabab Kreidieh Ward, M.P. Beddoes, Murray R. Hodgson, Bryan Gick · The Journal of the Acoustical Society of America · 2005
Estimating glottal waves by inverse filtering vowel sounds and deriving vocal-tract area functions (VTAFs) from vocal-tract filter (VTF) estimates require that VTF models be realistic and that VTF estimates contain no effects of open glottises and glottal waves. In this study, VTFs are modeled to have lip reflection coefficients with low-pass frequency responses; to minimize the effects of open glottises and glottal waves on the estimates, VTFs are estimated from sustained vowel sounds over closed glottal phases, assuming that the glottal waves are periodically stationary random processes. Since incomplete glottal closures are common, VTF estimates may contain the effects of glottal loss. To eliminate the effects of glottal loss in the VTF estimates, lip-opening areas must be known. Theoretically, estimates of glottal waves and VTAFs corresponding to large-lip-opening vowel sounds are less affected by the glottal loss than those corresponding to small-lip-opening vowel sounds. The VTAFs and glottal waves estimated from vowel sounds produced by several subjects are presented. The normalized VTAFs estimated from large-lip-opening sounds are similar to that measured from an unknown subjects magnetic resonance image. Over closed glottal phases, the glottal waves are non-zero. They increase during vocal-fold colliding, and decrease or even increase during vocal-fold parting.