Events Calendar

Detecting paralinguistic information from speech and language for clinical applications: Algorithms and information limits

Speaker: Visar Berisha, Arizona State University

Talk Title: Detecting paralinguistic information from speech and language for clinical applications: Algorithms and information limits

Abstract: The ability to share our thoughts and ideas through spoken
communication is fragile. Even the simplest verbal response requires a
complex sequence of events. It requires thinking of the words that best
convey your message; sequencing these words appropriately; and then
sending signals to the muscles required to produce speech. The slightest
damage to the brain areas that orchestrate these events can manifest in
speech and language problems. These disturbances offer a window into
brain functioning. In the first part of this presentation, I will
present an overview of a number of projects where we use interpretable
measures of speech and language production as proxies for cognitive and
motor health. The algorithms behind this work have practical utility in
clinical applications and can help answer basic research questions
related to dysarthric speech production.

In the second part of the talk, I will discuss new results from
non-parametric statistical signal processing that allow us to
characterize the information limits in speech. In contrast to existing
methods based on machine learning, this work provides a framework to
answer fundamental questions such as 'What are the bounds on how well I
can recover a parameter of interest from speech?' or 'How well should an
optimally trained classifier work for a particular application?'

Biography: Visar Berisha is an Assistant Professor at Arizona State
University with a joint appointment in the School of Electrical Computer
and Energy Engineering and the Department of Speech and Hearing Science.
Prior to joining ASU, Berisha was a research scientist at MIT Lincoln
Laboratory and then Principal Research Engineer for a Fortune 500
company. His research interests include speech analytics, statistical
signal processing, and information theory. Much of his recent work spans
all three of these fields to answer basic questions related to the
limits of information in speech. His research has led to many academic
publications, several licensed patents, and a revenue-positive startup
company. Berisha's work has been featured in the Science section of the
New York Times, on National Public Radio, and a number of other national
media outlets.

Host: Shrikanth Narayanan

Location: Hughes Aircraft Electrical Engineering Center (EEB) - 132

Audiences: Everyone Is Invited

Contact: Tanya Acevedo-Lam/EE-Systems