The challenges of developing articulatory synthesis models of early vocal production in humans

Abstract

At birth, human infants can cry, but they cannot articulate anything like the sound patterns of their mothers' speech that they have been hearing in the last month or so in the womb. Learning to speak involves a re-tuning of the muscle systems for breathing, crying, and suckling in order to be able to articulate a rapid sequence of exquisitely coordinated movements of the speech articulators—gestures of the larynx, velum, tongue, and lips—in synchrony with a lengthened expiratory phase of the breathing cycle, to acoustically shape the air flowing from the lungs to produce the sound patterns that are the words of a specific language. Models of speech production developed using data on articulator movement and coordination in adults cannot be applied directly, because this learning takes place in the context of substantial changes to the size and shape of the vocal tract over the course of physical maturation. This talk describes the changes in vocal tract morphology in human development, and then reviews how articulatory synthesis systems developed for adults have been adapted to model the expanding repertoire of vocalizations that have been observed in infants with normal hearing over the first year of life.At birth, human infants can cry, but they cannot articulate anything like the sound patterns of their mothers' speech that they have been hearing in the last month or so in the womb. Learning to speak involves a re-tuning of the muscle systems for breathing, crying, and suckling in order to be able to articulate a rapid sequence of exquisitely coordinated movements of the speech articulators—gestures of the larynx, velum, tongue, and lips—in synchrony with a lengthened expiratory phase of the breathing cycle, to acoustically shape the air flowing from the lungs to produce the sound patterns that are the words of a specific language. Models of speech production developed using data on articulator movement and coordination in adults cannot be applied directly, because this learning takes place in the context of substantial changes to the size and shape of the vocal tract over the course of physical maturation. This talk describes the changes in vocal tract morphology in human development, and then reviews h...