Application of Voice Analysis Programs to Quantify Fatigue
Department
Chemistry
Major
Chemistry
Research Advisor
Thimgan, Matthew S.
Advisor's Department
Biological Sciences
Funding Source
Missouri S&T OURE Fellows, BIC-Missouri S&T College of Arts, Sciences, and Business
Abstract
Sleepy people exhibit increased cognitive impairment, a higher likelihood of falling asleep involuntarily, and increased error rates. These fatigue related effects are concerning for fields like healthcare and transportation. It would be useful to have a fast, reliable, inexpensive, and objective method to quantify the sleepiness level of an individual.
Commonly, subjective surveys and cognitive tasks are used to assess sleepiness, but these methods are either flawed or prohibitively time consuming. Previously, a person’s speech patterns have been shown to indicate fatigue level, and this project sought to apply this methodology as an objective readout of sleepiness in a real-world situation.
Human subjects completed both subjective and objective tasks to characterize them as sleepy or not. Mel-Frequency Cepstrum transformations of voice samples were used with Hidden Markov Modeling to build fatigue prediction models. The models were then evaluated to categorize people as sleepy or alert based on their sleep patterns.
Biography
Robert Block is a Senior studying Chemistry with a Pre-medicine emphasis and a minor in Biological Sciences. After graduation he plans attending Pharmacy school. He has had internships at Vanderbilt University Medical Center, where he completed data visualization and delivery projects with the VAPIR team. He has also had several research projects at Missouri S&T, including work with drosophila flies and Nuclear Magnetic Resonance Spectroscopy. Robert mentors the Rolla High School FIRST Tech Challenge Robotics teams. He enjoys shooting sports, canoeing, programming, martial arts, and tinkering.
Presentation Type
OURE Fellows Final Oral Presentation
Document Type
Presentation
Location
Carver Room
Presentation Date
11 Apr 2017, 1:00 pm - 1:30 pm
Application of Voice Analysis Programs to Quantify Fatigue
Carver Room
Sleepy people exhibit increased cognitive impairment, a higher likelihood of falling asleep involuntarily, and increased error rates. These fatigue related effects are concerning for fields like healthcare and transportation. It would be useful to have a fast, reliable, inexpensive, and objective method to quantify the sleepiness level of an individual.
Commonly, subjective surveys and cognitive tasks are used to assess sleepiness, but these methods are either flawed or prohibitively time consuming. Previously, a person’s speech patterns have been shown to indicate fatigue level, and this project sought to apply this methodology as an objective readout of sleepiness in a real-world situation.
Human subjects completed both subjective and objective tasks to characterize them as sleepy or not. Mel-Frequency Cepstrum transformations of voice samples were used with Hidden Markov Modeling to build fatigue prediction models. The models were then evaluated to categorize people as sleepy or alert based on their sleep patterns.
