On acoustic salience of vowels and consonants predicted from articulatory models

National audienceThis presentation deals on acoustic salience of vowels and consonants predicted from articulatory modeling (AM). An articulatory model describes vocal tract (VT) profiles with a small number of parameters that are interpretable in phonetic terms. Such a modeling allows a better understanding of the relationship between articulatory configurations and acoustic patterns. The goal of this paper is twofold. The first part is rather an introduction to the issue and aims to illustrate AM. The advantages and limits of studying speech with different types of modeling, such as simple models of VT area function and more elaborated AM, are very briefly reviewed, and articulatory models are introduced. In anthropological Maeda's model, the area function is reconstructed with seven articulatory parameters: one for the jaw position, three for the tongue, two for the lips and one for larynx height. Such an AM makes it possible to evaluate the separate contribution of individual parameters to the vowel formants, and their interaction. In the second part, I propose some hints to complement the traditional way of transcribing the sounds and coarticulatory effects by acoustic and articulatory notation. It is based on the notion of the Fpattern defined by Fant (1960) and the use of Maeda's model to explore the maximal articulatory and F-pattern spaces. Such exploration reveals the similarities of the F-pattern among vowels, semi-vowels, and consonants that share a similar tongue position and shape and lip configuration, but differing in the constriction size and shape. Exploring the whole F-pattern space helps to define prototypical vowels specified mainly by specific acoustic characteristics, and which may serve as references. This approach makes it possible to describe subtle but audible and consistent differences between the realizations of a similar phoneme of different languages traditionally represented by the same IPA symbols. Moreover, it incorporates coarticulatory and positional influences therefore thus uncovering the seeds for perceptual confusions among sounds due to their immediate phonetic context, which in turn can cause sound changes