FUNCTIONAL DATA ANALYSIS OF ARTICULATORY VARIABILITY IN VCV SEQUENCES

ABSTRACT: We use Functional Data Analysis (FDA) to examine articulatory variability across repetitions in normal speech, under different movement constraints. The data base consists of horizontal and vertical articulatory movements of lips, jaw, and tongue in vowel-consonant-vowel sequences, recorded using a magnetometer. FDA is applied to temporally normalize and align the trajectories, and visualize the distribution of temporal and amplitude variability along the trajectory patterns. For quantitative assessments, we also define and compute an index of amplitude variability, defined as the mean standard deviation between peak velocities of the consonantal closure by the active articulator, in each VCV sequence. Our results suggest that variability differs both across articulators and segment types in ways that probably reflect task-specific production constraints. We conclude that articulatory variability in normal adults varies as a function of both the phonetic requirements of the consonant and the biomechanical characteristics of the articulatory structures involved