Glottal inverse filtering of nasalized vowels
Paul Milenkovic, Feng Mo · The Journal of the Acoustical Society of America · 1986
Linear prediction provides a means for automatically determining an inverse filter for estimating the glottal volume velocity waveform from the acoustic speech waveform. The behavior of this type of glottographic analysis is considered for nasalized vowels. The analysis is applied to non-nasal and nasal productions that are synthesized using a transmission line model of the vocal and nasal tracts. The analysis is also applied to nasal and non-nasal vowel productions made by a male speaker. The effect of the nasalization is to introduce extraneous ripple onto the closed phase portion of the glottal volume velocity waveform without changing the observed open quotient. This ripple results from a closely spaced resonance antiresonance pair that is not compensated by the all-zero linear prediction inverse filter. The low level of the ripple is accounted for by the narrow frequency range over which the inverse filter is in error. [Work supported by NIH.]