Undergraduate students contribute to research projects all the time at St. Thomas; for engineering senior John Fetzner, that research also has strong personal meaning. Last year, Fetzner approached engineering assistant professor Cheol-Hong Min because he was struck by Min's research topic: Developing an audio and motion sensor system for children with autism that can objectively measure what emotions they’re expressing.
“Most of my family, including myself, has moderate to severe hearing loss. … Even though it’s a different disability for me, this project is the same goal: working to help communicate with those who it’s much harder for,” Fetzner said.
The project is an extension of one Min started in graduate school, and is rooted in feedback from doctors and parents. Many have discussed the difficulty of tracking treatment progress based only on the subjective evaluations of parents, caretakers or teachers. A system is being developed that can track, with sensors, the behaviors and motions of nonverbal children with autism. The sound and video can then be analyzed and aligned with knowledge of what emotions are being communicated, resulting in an objective viewpoint of behavior to inform treatment.
“The ultimate goal is also real-time feedback. Say, when the child is making some noise it would alert the parent to know they’re not happy. Otherwise you have to be with the child 24/7,” Min said. “If you’re at a remote location, first floor of the house and the child is [on the] second floor, you could be provided through your phone or a wearable device an alert [of what emotions they’re expressing].”
Something as complex as creating an emotional map requires the development of an accurate algorithm, which is what Fetzner is working on full time this summer. He is breaking down video and audio of children with autism. With each inputted example, he increases the accuracy of the algorithm.
“It is satisfying to see [Min's] algorithm work for detecting stimming [repeated physical movements associated with people with autism] without human intervention. The goal is to eventually not only use the recognition of the stimming to determine if it’s a positive or negative emotion, but potentially break those out into different emotions like angry or sad, or with the positive side to happy and satisfied,” Fetzner said. “[Improving the algorithm] is like if we gave you a set of words you didn’t know and had a dictionary to look it up, and each time you get a new word you can put it in your dictionary and add it, to make the dictionary better. … [The algorithm] gets better at what it’s doing.”
In the process, Fetzner also gets better at what he’s doing: gaining experience in a career field where he can help make the world better.
“Not only is it great research, but it fits perfectly in the field I’d like to go into working with communication, and it really worked toward my goal of trying to help people,” he said. “It’s easy money for engineers to go into oil or something else that’s less directly helping people, but I’ve always hoped I could go into the medical field or something similar where I’m developing solutions to help make people’s lives easier.”
