INTERDISCIPLINARY APPROACHES FOR ADVANCING ARTICULATORY SPEECH THEORY AND SYNTHESIS

摘要

Articulatory synthesis research has long been dominated by frequency domain and concatenate samplebased speech synthesis techniques. While successful in some domains (e.g., voice-based databases), these techniques still cannot produce natural looking and sounding speech from text for an arbitrary speaker. Natural looking and sounding speech technology is one of the next major milestones in voice-based interaction for natural user interfaces. Through a team of interdisciplinary international researchers, we have been steadily working towards creating the necessary platform to overcome basic problems in synthetic speech production. Our approach is three fold: 1. Create a Functional Reference for Anatomical Knowledge (FRANK) template model for use in studying articulatory synthesis, 2. Develop new methods for voice synthesis that couple to the platform, and 3. Use the biomechanically driven articulatory synthesizer to explore and develop theories of speech production. Our team draws from electrical, computer and mechanical engineering, linguistics, computer science, dentistry, radiology and surgery to assemble the needed components for advancing the state of the art.