Toward phonetically grounded

The purpose of this paper is to correlate the four formant frequencies F1–F4 and quality factors Q1–Q4 with the positions, areas, or area ratios formed by the four active articulators: tongue root, tongue body, blade, and lips. The quality or amplification factor Q is the formant frequency F divided by the bandwidth B. Toward this end, a 27-tube frequency-domain vocal tract model (FDVT) is developed. Four articulator regions are delimited in the model: an 8-tube tongue root region, a 9tube tongue body region (one-quarter wavelength of F2), a 6-tube blade region (one-quarter wavelength of F3), and a 4-tube lip region. Vowel area functions of ten speakers were taken from seven X-ray and MRI studies and fit to 27 equal-length tubes using cubic spline interpolation. Correlation matrices between the acoustic and articulatory parameters are calculated for the vowel system of each speaker. The coefficients of the parameter pairs are averaged across the ten speakers and the most highly correlated ones are ranked. Tongue root aperture (tongue root area normalized by lip area) is shown to be inversely correlated with F1 frequency. When the tongue body and blade constrictions move toward the lips through their respective one-quarter wavelengths, the F2 and F3 frequencies also shift higher. Tongue body aperture (tongue body area normalized by lip area) is directly correlated with Q2. Similarly, blade aperture (blade area normalized by lip area) is directly correlated with Q3. Lip protrusion displays an inverse correlation with F4 frequency. Lip aperture (lip area) has a moderate direct correlation with F1 and a weaker inverse correlation with Q4. In agreement with perturbation analysis, F1 frequency is found to be closely correlated with the tongue root area, but negligibly correlated with the tongue body area. Consequently, F1 distinctions among high, mid, and low vowels are made by varying the tongue root aperture and not, as is traditionally assumed, by raising or lowering the tongue body.