Human and Machine Music Perception at CCRMA Hearing Seminar

Date: 
Thu, 02/18/2010 - 11:30am
Location: 
Seminar Room at the Knoll
Event Type: 
Hearing Seminar

New time: 11:30 instead of 11!!!!!

I'm really happy to introduce Luke Barrington to the CCRMA community.  Luke is from UCSD and will be visiting on February 18th.  He will lead the discussion at the Hearing Seminar about his work on human and machine music perception.

Luke is a founding member of the UCSD computation audition lab.  This lab has been doing many very creative and smart things with music, both to build machines that understand music and to design tests that elicit human reactions to music.  Luke and his colleagues have built some nice games the entice people to label music, and have used this data in several innovative ways to build machine-learning system (or should I say music-learning systems.)

    Who:    Luke Barrington (UCSD)
    What:    Automatic Music Perception
    When:    Thursday, February 18th at 11AM
    Where:    CCRMA Seminar Room (behind the elevator on the top floor)

Bring your musical ears to CCRMA on Thursday morning.

- Malcolm
P.S.  Just a reminder to those of you that aren't on campus every day.  Parking is by permit only at the Knoll.  Best bet is to come early and park in the Tressidor lot and walk up the knoll to CCRMA.




================ ABSTRACT ================
Machine understanding of music presents challenges across a variety of disciplines.  We use signal processing to capture the content of audio waveforms and online data mining to associate the music with descriptive semantics.  We develop machine learning models to make sense of all this information and apply learned knowledge to millions of new songs.

In this talk, I will present a start to finish approach for automating music understanding.  We begin by developing an online human computation game called Herd It (www.herdit.org) to collect semantic and social information about music.  The tags collected by Herd It are used to train supervised machine learning models that can generalize these labels to any song.  This semantic representation powers applications including music similarity and recommendation as well as a natural language music discovery engine that allows listeners to find the perfect tune without knowing song or artist names.
==========================================


=================== BIO ===================
Luke Barrington received the BE (elec) degree from University College Dublin (UCD), Ireland in 2001 and MS degree from the University of California, San Diego (UCSD) in 2004.  He is about to complete his Ph.D thesis on "Machines that Understand Music" in UCSD's Electrical and Computer Engineering Department.  He is a co-founder of the UCSD Computer Audition Laboratory where he applies machine learning, signal processing, data mining and human computation to automating music understanding. 

In 2001, he was UCD's Young Engineer of the year.  In 2005 he was a National Science Foundation (NSF) EAPSI fellow in Japan.  From 2006 to 2008, he was the recipient of a US NSF IGERT Fellowship.  He is an avid musician and wails on the guitar.

Website: van.ucsd.edu
==========================================

Open to the Public
Syndicate content