During the period of time a child acquires the capacity to produce speech, their anatomical structure is continually developing and growing. Consequently, the characteristics of the speech sounds they produce also change over time, and are considerably different from those of an adult. Learning to produce intelligible speech sounds would seem to be a daunting task for a child. Nonetheless, normally-developing children acquire the ability to produce most speech sounds by age six. The overall aim of the proposed research is to better understand how children accomplish this remarkable feat. The researher's approach is to investigate the mechanisms of children's speech by developing a computer-based model of the trachea, larynx, throat, mouth, and nasal passages that will be configured to represent children 2 to 6 years of age. This model is capable of producing realistic simulations of vowels, syllables, words and sentences. These simulated speech samples can be studied with respect to how the structure of the child-like system influences the generation of sound waves and how those waves are shaped into the sounds of speech. The simulations will also be compared to a database of audio recordings of the speech of similarly aged children.
The common view of sound production for speech over the last 50 years has been largely based on what is known about the adult male as a talker. This view and associated theory has been remarkably successful as a description of adult male speech. It is the prevailing description in most speech science textbooks and is the basis of many types of speech synthesis and speech analysis systems in use today. This research project will challenge this view by proposing that the speech sounds of children are generated with mechanisms that are fundamentally different from adults. Results of this research project will enhance and expand theories of speech production and speech development in children, and will contribute to education in the area of speech science. In addition, the project will impact the next generation of systems for speech analysis, synthesis and recognition.
