Real-time feedback of speech movements based on optopalatography

Speech is the product of articulation and information on the sound forming articulatory gestures of the tongue and the lips is of major interest not only in speech therapy and phonetic research, but in speech synthesis and coding as well. Various attempts have been made to describe these important articulators using established imaging techniques like ultrasound, electromagnetic articulography (EMA), or X-ray microbeam [3], but so far none of these has proven to be fast, precise, and at the same time cheap enough to yield satisfying results that can be applied in day-today applications. However, the relatively little-known glossometry [2, 4], also called optopalatography (OPG) [6], has shown some good results in measuring the tongue contour in the anterior oral cavity [1] and holds the most promise as a clinically relevant tool as it is relatively cheap to produce, easy to employ, and reasonably precise. Its biggest limitation, though, is that its coverage does not extend to the large part of the tongue in the posterior region of the oral cavity. To remedy this drawback, we introduce in this poster an animated 2D articulation model controlled by real-time optopalatographic measurements of the positions of the upper lip and the tongue in the anterior oral cavity that are extended by linear prediction of the posterior part of the tongue. The measurement system is an improvement on a previous prototype with increased spatial resolution and an enhanced close-range behavior. The prediction coefficients were determined and evaluated using a corpus of vocal tract traces of 25 sustained phonemes. The model represents the tongue motion and the lip opening physiologically plausible during articulation in real-time. Finally we discuss how the model could be improved by speaker adaptation or by incorporating additional data like electropalatographic (EPG) measurements in our ongoing project to develop a 3D real-time representation of the vocal tract [5].