Engineering, music theory professor explains how computers see music

University of Rochester Professor Mark F. Bocko, right, talks with music tech entrepreneur Ryan Wrenn after today's presentation. (Heidi Hall/Vanderbilt University)

The sound of Ray Charles crooning “You Don’t Know Me” has been raising goose bumps on the heartbroken for five decades.

But what makes it bring a tear to the eye when the same tune produced on a MIDI doesn’t? And could a computer ever tell the difference in Charles’ famed version and those of Elvis Presley, Willie Nelson or Asleep at the Wheel’s Ray Benson?

Mark F. Bocko, a University of Rochester researcher who combines his computer engineering and music theory backgrounds, is using motif discovery in sequential data to answer those questions. He’s borrowing from methods of finding patterns in genes.

He discussed how it works – demonstrating his points with a series of sound clips, spectrographs and grids – in a lecture this afternoon in Jacobs Believed in Me Auditorium at Vanderbilt University School of Engineering.

His work “is like looking for a needle in a haystack, only you don’t know you’re looking for a needle,” he quipped.

“What makes music musical? From the engineering standpoint, it’s just a signal. It’s a sound wave propagating the air.”

The presentation made clear that Bocko himself sees it as much more than that.

He named six elements that separate musical performances — timing, articulation, vibrato, timbre, auditory loudness and pitch deviation – that could be searched for sequences.

He’s also researching how computers can graph individual notes. There’s already software that does it with 95 percent accuracy for one-note-at-a-time instruments, such as clarinets, Bocko said. But no software can sort out complicated, layered pieces, such as Sammy Nestico’s big band hit “Night Flight,” where horns, drums and bass play together.

Another project for Bocko, an electrical and computer engineering professor at the University of Rochester who also teaches at its music school, is his “adaptive dereverberation filter.” It would measure the echoes in a room and undo them for musical recordings – technology that also has drawn interest from the Defense Department for voice recordings, he said.

Bocko ended by saying a music-tech-producing marriage between the School of Engineering and Nashville’s Music Row would benefit both.

Contact

Heidi Hall, (615) 322-6614
Heidi.Hall@Vanderbilt.edu
On Twitter @VUEngineering