Rudolph Sock
Institut de la Communication Parlee, Grenoble, France
Anders Lofqvist
Haskins Labs., 270 Crown St., New Haven, CT 06511
The temporal organization of lip and jaw movements was investigated for one subject, producing six pseudowords. The words, produced at two speech rates, normal and fast, were: /pipi/, /papa/, /pupu/, /bibi/, /baba/, /bubu/. Movement data from the upper and lower lips, and the jaw were obtained via an optoelectronic tracking device, together with an audio recording. Based on articulatory and acoustic events in the velocity and audio signals, respectively, four cycles were identified in the movement of each articulator, and two cycles in the acoustic waveform. The movement cycles are: (1) maximum vocalic velocity; (2) vocalic cycle onset; (3) maximum consonantal velocity; and (4) consonantal cycle onset. Velocity cycles were determined, as the interval between successive positive or negative peaks associated with the raising or lowering gesture in the production of either a vowel or a consonant. Onset cycles were defined as the recurrence of zero-crossings on velocity traces, reflecting the onset of a lowering or a raising gesture to produce a vowel or a consonant. The acoustic cycles are: (1) release cycle; and (2) closure cycle. Vocalic and consonantal phases were defined, within vocalic and consonantal movement cycles, respectively, as specific timing relations of articulator lowering or raising gestures, and within the appropriate acoustic cycle, as the interval that presents a stable formant structure for the vocalic phase, and as the obstruent portion, for the consonantal phase. Gestural and acoustic phasing patterns obtained within the cycles suggest that: (1) the lower lip is an efficient articulator in distinguishing the front vowels /i, a/ from the back vowel /u/, in both consonantal contexts and in all four cycles; (2) the upper lip distinguishes all vowel types in both consonantal contexts in the vocalic cycle onset, and voicing contrast in the maximum consonantal velocity; (3) distinct phasing patterns in normal speech tend to merge with increased speech rate; (4) the jaw, although a carrier articulator, does not contribute actively to separating the phonetic tasks required in this study; (5) the closure cycle is the better of the two acoustic cycles in separating vocalic and consonantal classes. Results are analyzed and discussed in terms of cognitive tasks and emergent properties of the speech production system. [Work supported by the Fyssen Foundation and NIH.]