Tongues and lips without jaws: a comparison of methods for decoupling speech movements.

Speech-related motions of small markers attached to the tongue, lower lip, and lower jaw of 44 normal young adult talkers of American English were analyzed to estimate the relative accuracy of selected methods for decoupling tongue and lip motions from ongoing jaw motion when all movements are originally measured relative to a common reference frame (such as the head). In general, results of the analysis show that a common "simple subtraction" method that ignores pitching rotation of the jaw yields larger errors in positions and speeds of decoupled tongue and lip markers than methods that do not ignore rotation. When the jaw is widely opened, is moving quickly, and/or when any decoupled marker is relatively far from the origin of a local coordinate system fixed to the jaw, positional and speed errors associated with the subtraction method can exceed 5 mm and 25 mm/s, respectively. We propose a simple procedure for estimating the pitching rotation of the jaw that can be applied when pitch cannot be measured. We then show that decoupled motions of tongue and lower lip markers based on estimated jaw rotation involve less error than those derived from any other decoupling method considered. Careful attention to processing methods should yield more accurate inferences about the nature and degree of coordination between speech-related movements of the tongue and lower lip that are decoupled from, and hence independent of, concurrent movements of the lower jaw.