Harmony Between AI and Violin Instruction?

When much of the world moved online during the pandemic, so did the lessons taught by concert violinist Irina Muresanu—and it didn’t play out beautifully.

The violin and strings associate professor could easily assess bow technique and pose in person, but her ability to guide a student declined across a glitchy video chat connection.

Now, with a team of UMD experts, Muresanu seeks to bridge instructional gaps by creating a platform that combines violin teaching with advanced computing and artificial intelligence, all based on 350 years of violin tradition. While a student plays, their phone or other device “watches” them and provides feedback on form and sound while suggesting appropriate practice materials.

“As a teacher you really learn to become an expert at figuring out what teaching materials a particular student might need,” Muresanu says. “The idea of developing a platform that would actually offer this personalized feedback really resonates with me.”

Muresanu teamed up in 2021 with Cornelia Fermüller, a research scientist at the University of Maryland Institute for Advanced Computer Studies with expertise in computer vision and signal processing. Later, electrical and computer engineering Professor Shihab Shamma, a computational neuroscientist who specializes in sound perception, joined the team as well.

To help the app “see,” researchers analyzed violinists’ movements, recording video of participants ages 5 through 23; all played the same song on the same violin to ensure consistency.

The project received $900,000 from the National Science Foundation in August, along with funding from UMD’s $30 million Grand Challenges Grants program last year. After developing the app’s user interface, the group hopes to release it in the next two years to expand opportunities for people who can’t afford private lessons or connect with an expert instructor—at least a human one.

“This project aims to make instrumental music education available for all,” Fermüller says.