A Noninvasive Acoustic Method Using Frequency Perturbations and Computer-Generated Vocal-Tract Shapes

This study investigated improved processing of acoustic data with two adult Down's syndrome subjects. Sustained vowel samples were processed through a fast-Fourier-transform spectrum analyzer, and digital waveform data were used to obtain period-by-period measurements of the fundamental frequencies. Unusual frequency perturbation (jitter), later identified as diplophonia, was found for one of the Drain's subjects. In addition, the first three formant frequencies of the vowels were determined and, utilizing an algorithm described by Ladefoged and his colleagues, computer-generated vocal-tract shapes were plotted. Differences in vocal-tract shapes, especially for the back vowels, were observed between the Down's female and the normal shape. Correlations between vocal-tract shapes of the Down's subjects and those for a normal man or woman were computed. A partial three-way factor analysis was carried out to determine those lead factors or coefficients for each subject that were due to individual differences. These procedures, offering synthesized techniques portraying the interparingeal/oral functioning of the speech structures, may eventually have direct noninvasive diagnostic and therapeutic benefit for voice/resonance-disordered clients.