Quantitative Measures of Intelligibility of Dysarthric Speech

Project: Research project

Project Details


Reduced intelligibility is the at the heart of the communication disorder associated with the dysarthrias and other speech production deficits, causing functional and societal limitations and disabilities, and undermining quality of life. At present, the most common clinical approach to intelligibility assessment is subjective in nature. Subjective impressions, although expedient to collect, are by their very nature vulnerable to clinician bias. Safeguards to enhance reliability and validity of subjective impressions (e.g., nave listeners, listen panels, etc) are prohibitively high cost for typical clinical settings. There is an urgent need for the development of dependent measures sensitive enough to detect speech changes in early disease; document the efficacy of behavioral, surgical or pharmacological interventions; and track speech changes associated with disease progression. In this translational research, we propose to use the well-developed principles of intelligibility quantification that have been standardized in speech telecommunications and to develop and validate new measures of intelligibility in clinical populations. Specifically, we aim to develop a process by which the intelligibility of dysarthric speech can be validly and reliably quantified relative to a reference control sample; and to algorithmically isolate the source(s) of intelligibility degradation along certain dimensions to augment clinical practice. In addition to integrating existing intelligibility metrics within our developed framework, we intend to instantiate new metrics, specifically optimized for dysarthric speech, based on insight and data gained during a previous NIH-funded grant. Preliminary results show promise that the proposed approach will yield a successful framework for quantifying intelligibility. These results indicate existing telecommunication-based intelligibility metrics accurately capture intelligibility degradation for noisy speech. Further, we show that the same metrics can accurately capture differences in intelligibility between three dysarthric speakers of differing intelligibility levels. Finally, we show that a subset of the envelope modulation spectra features, developed previously in our lab, are remarkably predictive of a speakers intelligibility. The results of the proposed research will form the basis for a subsequent R01 proposal for the development of a fully functional clinical tool to objectively quantify and track speech intelligibility and its causal components.
Effective start/end date8/1/127/31/15


  • HHS: National Institutes of Health (NIH): $392,031.00

Fingerprint Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.