A mathematics professor is developing a method that employs artificial intelligence to clearly understand the electrical brain activity data conveyed through electroencephalogram monitoring.
Vasileios Maroulas’s method has applications in neuroscience and is of interest to the US Army Research Laboratory’s Brain-Computer Interface initiative. The Army Research Office has funded Maroulas’s work since 2017. Maroulas was appointed as a senior research fellow at the US Army Research Laboratory in September.
“We are trying to identify patterns of brain activity in people who can make the right decisions instantaneously,” Maroulas said.
The Brain-Computer Interface project aims to develop technologies for recording brain activity and establish computational methods to translate the signals into computer-executable commands. Using Maroulas’s method, Army researchers hope to harness the complex information collected from a soldier’s neuroimaging data.
Physiological signal data—such as that collected by EEG, electrocardiograms, electromyograms, and electrooculograms—contains a lot of nonessential information that can obscure the relevant data. To bypass this extraneous “noise,” Maroulas is developing a statistical framework for topological data analysis (TDA) to extract the meaningful information.
“The TDA algorithm will likely help enable future Army applications in artificial intelligence that could aid the soldier and commander in target detection, predictive movement, and go or no-go decision making,” said Joseph Myers, mathematical sciences division chief in the Army Research Office, an element of U.S. Army Combat Capabilities Development Command’s Army Research Laboratory. “The algorithm will also help address data extraction and identification in medical imaging for traumatic brain injuries, signal processing for object and facial recognition, and in data storage and analysis.”
“We’re trying to discover shape patterns in data,” Maroulas said. “Although there are several techniques out there, we are going a totally different route where we merge statistics and machine learning with topology and geometry.”
—
CONTACT:
Karen Dunlap (865-974-8674, kdunlap6@utk.edu)
Amanda Womac (865-974-2992, awomac1@utk.edu)