Michel Pitermann

Effect of speaking rate and contrastive stress on formant dynamics and vowel perception

Vowel formants play an important role in speech theories and applications; however, the same formant values measured for the steady-state part of a vowel can correspond to different vowel categories. Experimental evidence indicates that dynamic information can also contribute to vowel characterization. Hence, dynamically modeling formant transitions may lead to quantitatively testable predictions in vowel categorization. Because the articulatory strategy used to manage different speaking rates and contrastive stress may depend on speaker and situation, the parameter values of a dynamic formant model may vary with speaking rate and stress. In most experiments speaking rate is rarely controlled, only two or three rates are tested, and most corpora contain just a few repetitions of each item. As a consequence, the dependence of dynamic models on those factors is difficult to gauge. This article presents a study of 2300 [iai] or [i epsilon i] stimuli produced by two speakers at nine or ten speaking rates in a carrier sentence for two contrastive stress patterns. The corpus was perceptually evaluated by naive listeners. Formant frequencies were measured during the steady-state parts of the stimuli, and the formant transitions were dynamically and kinematically modeled. The results indicate that (1) the corpus was characterized by a contextual assimilation instead of a centralization effect; (2) dynamic or kinematic modeling was equivalent as far as the analysis of the model parameters was concerned; (3) the dependence of the model parameter estimates on speaking rate and stress suggests that the formant transitions were sharper for high speaking rate, but no consistent trend was found for contrastive stress; (4) the formant frequencies measured in the steady-state parts of the vowels were sufficient to explain the perceptual results while the dynamic parameters of the models were not.

An inverse dynamics approach to face animation

Muscle-based models of the human face produce high quality animation but rely on recorded muscle activity signals or synthetic muscle signals that are often derived by trial and error. This paper presents a dynamic inversion of a muscle-based model (Lucero and Munhall, 1999) that permits the animation to be created from kinematic recordings of facial movements. Using a nonlinear optimizer (Powells algorithm), the inversion produces a muscle activity set for seven muscles in the lower face that minimize the root mean square error between kinematic data recorded with OPTOTRAK and the corresponding nodes of the modeled facial mesh. This inverted muscle activity is then used to animate the facial model. In three tests of the inversion, strong correlations were observed for kinematics produced from synthetic muscle activity, for OPTOTRAK kinematics recorded from a talker for whom the facial model is morphologically adapted and finally for another talker with the model morphology adapted to a different individual. The correspondence between the animation kinematics and the three-dimensional OPTOTRAK data are very good and the animation is of high quality. Because the kinematic to electromyography (EMG) inversion is ill posed, there is no relation between the actual EMG and the inverted EMG. The overall redundancy of the motor system means that many different EMG patterns can produce the same kinematic output.

Experimental study of the target undershoot model using acoustic and ‘‘articulatory’’ data

In a previous study [M. Pitermann and J. Schoentgen, 1st ESCA Tutorial and Research Workshop on Speech Production Modeling, 17–20 (1996)], the influence of speaking rate and emphatic stress on formant frequencies and corresponding target estimates were analyzed for the vowels [a] and [e] in an [i‐i] context. Unreached targets were estimated by means of two dynamic and two kinetic models of formant transitions. The results showed that both formant frequencies and target estimates varied with speaking rate and emphatic stress. In the present survey, we analyzed ‘‘articulatory’’ targets for the same corpus in the framework of a vocal tract model. The ‘‘articulatory’’ trajectories were analytically calculated by means of an acoustic‐to‐articulatory inversion carried out on a six‐tube Kelly‐Lochbaum model of the vocal tract [S. Ciocea, ‘‘Semi‐analytic formant‐to‐area mapping,’’ Ph.D. thesis (1997)]. Targets were then estimated by means of asymptotes or steady‐state solutions of transition models fitted to traj...