This project addresses ways in which speakers use the vocal tract to create speech (production) and listeners extract the underlying meaningful parts of the signal (perception). Speech is an essential part of human society, and great effort is put into learning to talk, speaking with others, and repairing difficulties in speech when necessary. Speech is also quite variable, and the variability is due to both structured and random factors. The structured variability provides information about, for example, the size of the speaker?s vocal tract, where the speaker comes from, and what speech segments are adjacent to the one being produced. The random variability is often seen as unhelpful, but recent work indicates that it can provide flexibility that is beneficial. In the current proposal, measurements of speech articulation and acoustics will be combined with mathematical modeling to distinguish flexibility from diminished control. Flexibility is also essential in perception, so that the wide range of speakers one encounters, with their individual speech characteristics, can still be understood. Flexibility around baseline control is important in production so that communication can be tuned to specific situations, such as speaking more clearly to someone who does not speak the language natively. The sources of variability in speech will be examined in production experiments where the tongue, lips and other articulators are measured with such approaches as ultrasound and electromagnetic articulometry, and in perception experiments, where the output of the vocal tract variability can be measured for its ability to convey a message. Mathematical methods that use nonlinear dynamics will separate the controlled and random variability present in the production signals. The results will provide a better foundation for assessing speech difficulties, such as second language accent or stuttering, and should suggest new means of remediation. In these studies, biofeedback will be used to show the speaker what their tongue is doing and what it should be doing. The topics explored?articulator movement, articulatory settings, and variability?are globally applicable to speech in any language. They are foundational issues and thus not likely to be fully understood from any one set of studies, but the current studies will considerably advance our knowledge and our ability to model that knowledge.
The proposed research addresses ways in which the vocal tract creates speech (production) and how listeners extract the underlying linguistically significant parts of the signal (perception), and the ways in which the two are linked. The research will use measurements and modeling to separate controlled from random variability. This will help to distinguish flexibility from lack of control, allowing the testing of new biofeedback techniques for speech remediation.
Showing the most recent 10 out of 72 publications
