Multi-Layer Nonlinear Mapping Networks for Determining Articulator Positions from Speech

In recent years, there has been a big upsurge in research on artificial neural networks. Although the primary motivation for this research comes from a desire to understand the principles underlying the extraordinary processing capabilities of the human nervous system, the nonlinear mapping properties of the neural networks by themselves are of great value in finding solutions to many difficult unsolved problems. We discuss in this paper the application of neural networks for determining the positions of different articulators (such as, the tongue, lips, velum, etc.) in the vocal tract from the acoustic information in the speech signal.