Vowel Kinematics in Various Consonant, Stress, and Phrase Position Contexts

Abstract:

Vowel kinematics were studied using both acoustic and articulatory data. Two sets of data were analyzed, primarily with Pearson correlation coefficients. For the first set of data, talkers were asked to produce four vowels in symmetric consonant context with labial and alveolar stops, in various phrase positions, and with and without emphasis. The second formant frequency trajectories of these vowels were fit with second-degree polynomials. Despite being requested to produce full vowels, even when they were not emphasized, the talkers tended to produce less extreme values of F2 as the duration of the vowel decreased. There were strong correlations between the quadratic and linear coefficients of the second-degree polynomial fits. These coefficients are proportional to the average acceleration throughout the F2 trajectory and the average velocity in the opening phase (from consonant release to F2 extreme), respectively. We endeavored to see whether the same relations held in the articulatory domain as well. For the second set of data, various syllables spoken by two talkers from the X-ray Microbeam Speech Production Database were examined. Multiple tokens of three syllables of one talker and one syllable for another talker were analyzed, where the consonantal contexts of the vowels were approximately symmetric. Second-degree polynomials were used to fit the F2 trajectories of the vowels. There was again a strong correlation between the average acceleration and average velocity for the F2 trajectories. Using the time where the F2 reached its extreme value as the landmark time, we calculated the distances of the pellets from this landmark time for each utterance. These distance trajectories were fit with second-degree polynomials, and the average acceleration and average velocity were found to be well-correlated. This usually remained true for polynomial fits taken over the open phase. Curve-fitting and correlation analysis can be useful for identifying underlying models for kinematics of speech articulators.