Machine Learning Models of the Tongue Shape during Speech

We describe our ongoing work on data-driven models of the tongue shape. Recording techniques such as EMA and X-ray microbeam track the position of 3–4 pellets on the tongue. Our models allow a realistic reconstruction of the full shape of the tongue with submillimetric accuracy from the location of these pellets, and rapid adaptation of an existing model trained with lots of data from one speaker to a new speaker for which little data is available. These reconstruction models are useful in several applications, such as to display the tongue in a talking head animation, to visualise the vocal tract in speech production, therapy and learning, to track more robustly a signal (such as the speech or the ultrasound image), or in articulatory inversion and synthesis.