This Project Grant award from the National Science Foundation Division of Behavioral and Cognitive Sciences provides $245,398 to support research estimating articulatory constriction place and timing from speech acoustics under the Social, Behavioral, and Economic Sciences federal grant program (CFDA 47.075).
The award funds development of a speech inversion tool to recover details of articulatory movements from acoustic speech recordings using machine learning methods. Researchers will extend an existing system's functionality to encompass nasality and voicing, and validate the expanded system using a new corpus of acoustic and articulatory data from American English speakers. Validation will also test the tool on productions from speakers of Canadian French and Russian.
When validated, the speech inversion tool is intended to help identify medical issues affecting speech movement organization and track changes from conditions like depression and schizophrenia. The ability to analyze articulatory movements from audio alone could substantially improve automatic speech recognition systems and assist various professionals studying speech under field conditions. This award supports the NSF's mission to promote basic research and education in the social, behavioral and economic sciences.
Generated 1/7/24, 10:38 AM