Alumni Magazine - Spring 2008 - (Page 53) a song and look at its fingerprint,” Chordia says. “We figure out what about it doesn’t change even when we have noise, even when we take a random section of it. Despite all of these transformations, what remains the same? “The essence of what we’re doing is called pattern recognition. We’re trying to find out what are the essential qualities and those things that really don’t change irrespective of all the different kinds of variations and permutations. … That’s what music information retrieval is all about.” Just the Beginning lark believes that the future of Georgia Tech’s music program lies in its interdisciplinary approach to performance and education. “We are not a program that is or should be bound in a disciplinary box or put in some institutional silo,” he says. “When I realized that we have every major on campus represented in our classes and ensembles, I saw that there was a connection between students’ interests and music. I couldn’t help but believe that some of those same interests were shared by faculty in other disciplines. That’s proven to be true, and I’ve seen more support for music by my colleagues at Georgia Tech than at any previous institution.” Clark envisions a myriad of possible collaborations with colleagues from across campus — from working with materials science researchers to improve the quality of a piano’s sound by replacing the traditional felt on its hammers with nanomaterials to working with College of Architecture and mechanical engineering professors to construct buildings with superior acoustics. He also sees interest in a new undergraduate degree in music and the potential for dual-degree programs, particularly to coincide with the College of Computing’s new threads platform. “I believe for music to succeed at Georgia Tech, we need to partner with every college on campus — we need to be looking at the major research themes and directions of the Institute. We are just at the beginning of that process and the possibilities are endless,” Clark says. “Five years ago, our facilities were in total disarray, the budget was a disaster, we were seriously understaffed and the idea of a degree of any kind in music seemed remote. Presently, there is an interdisciplinary graduate degree, performance opportunities have expanded, facilities continue to improve, community involvement is growing and we are actively pursuing commercial and entrepreneurial partnerships. I am extremely proud of what our students and faculty have accomplished and I can’t wait to see where we are five years from now.” GT “We have increasingly the ability to access any music at any time, anywhere,” says assistant professor Parag Chordia, who, with his students, is developing software to make sense of music. ping songs to their bare essentials to improve music recommendation applications. “What we’re trying to do is create machines that can listen and learn when they listen to music,” Chordia says. “It’s kind of similar to speech recognition except rather than working on speech, we’re working on music and trying to do things like extract melody or infer the beat or figure out what type of music it is, all for a variety of different types of applications. “We have increasingly the ability to access any music at any time, anywhere. So we have this type of situation where we’re inundated with musical information,” Chordia says. “People are increasingly interested in basically creating tools that will help them make sense of all of this music. “One of the approaches that we’re working on here is analyzing the actual content of the music to understand what makes this music tick. What is the real structure of this music? What are the characteristics that give it a certain feel? … We’re trying to create some kind of model that recreates how we as humans relate to music, and if we can successfully do that and start to build these models of songs, we may be able to say, ‘If you like this, I know other music you will like.’” Another music information retrieval application that may benefit from an improved song model is audio fingerprinting. If you’ve ever had a song stuck in your head for days but can’t quite place it, wouldn’t it be grand to simply hum a few bars into your cell phone and have the title sent to you in a text message? That’s now made possible through audio fingerprinting. “Audio fingerprinting has become very robust, and it relies on the same kinds of technologies, where we rip apart C Georgia Tech Alumni Magazine • Spring 2008 53
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.