Speech Production Research

A long-term research goal is to enable better understanding of the biological and physical basis of how humans produce speech and how they interact using spoken language and utilize such knowledge toward better speech engineering.

Specifically, two major goals of the research here are to directly investigate acoustic and perceptual consequences of articulatory changes and to derive realistic models for speech production and analysis. Fundamental questions that drive this work include:

• What makes human speech sound natural and what are its implications for machine synthesis of speech?
• What characterizes spontaneous speech, including affective qualities, and what are its implications in automatic speech recognition?
• How can we better explain the intra- and inter-speaker variability that is so common in speech?
• What are the articulatory-acoustic correlates of developmental speech production?

While the primary motivation of the research is geared toward developing medical and technological applications (such as devising objective assessment of disordered speech, creating natural-sounding speech synthesis and robust automatic speech recognition), it also provides a vehicle for validating and refining theories of human speech communication.

Research topics include: Physiologically-motivated modeling of human speech production, Imaging applications for speech research, Instrumentation including novel micro-sensors for direct measurements in the vocal tract, Acoustic-Phonetics, Articulatory methods in Automatic Speech Recognition and Speech synthesis.